This paper was produced to assist you in improving the performance of your OS/2 system by providing information on OS/2 Warp 4, its various subsystems, and practical performance tuning tips and techniques. It begins by familiarizing you the new features and enhanced functions of OS/2 Warp 4. It will then introduce you to the concepts and terminology needed to successfully tune an OS/2 system. You will learn to recognize the symptoms of many common performance problems. You will also learn how to create and use a repeatable process for analyzing a system's performance problem and understand the many tuning parameters and techniques used to fully optimize your system. The tuning tips and techniques are divided into 3 sections targeted to different levels of complexity: basic OS/2 tuning, advanced OS/2 tuning, and tuning tips for a networked OS/2 workstation.

This paper documents the procedures used and knowledge obtained from performance evaluation and measurement activities. Some procedures have been recorded in documents previously published by OS/2 development and support groups within IBM. It should be considered a working document subject to forthcoming changes without notice as additional information is obtained.

The information contained in this paper represents the interpretation of IBM on the issues discussed based on the measurement results as of the date of publication. Since the results are applicable only to the system under test, IBM makes no commitment nor guarantees the accuracy of any data measured after the date of publication.

The authors wish to express their sincere thanks to the comments, suggestions and reviews made by Michael Martino, Cyndi Kubich, Rich Wahl, Robert Paulsen, Jimmy DeWitt, Ivan Eisen, Bob Russell, Steve Woodward, Lannes Robinson, Valerie Jackson, and many others that have involved with OS/2 Warp products.

For further references, please refer to the following documents.

[1] OS/2 Warp 4 Readme

[2] OS/2 Warp 4 Reference and Commands

[3] OS/2 Warp 4 Tasks

[4] "Performance Tuning OS/2 Warp," Ron Cadima, ISV Development Assistance, IBM Boca Raton, FL, 1995.

[5] "OS/2 2.1 Performance Tuning for End Users," IBM, May 1993.

[6] "OS/2 Warp Performance Pentium Pro P6 Performance Benchmark Guide," Duc Vianney, IBM TR54.915.

[7] "OS/2 Client Tuning Worksheet," Tony White, IBM, Personal Systems Competency Center, May 1996.

The following terms are trademarks or registered trademark of the IBM Corporation in the United States or other countries

OS/2, OS/2 Warp, OS/2 Warp Connect, OS/2 Warp Server, OS/2 LAN Server, DB/2, LAN Server 4.0, Bonus Pak, Personal Communications/3270 , Visual Age and Visual Builder, ThinkPad, NetFinity, OpenDOC, WebExplorer, SOM/DSOM, DIVE, System Performance Monitor/2 (SPM/2)

The following terms are trademarks of other companies:

TRADEMARK OWNER

All other marks are the property of their respective companies.

References in this publication to IBM products, programs, or services do not imply that IBM intends to make these available in all countries in which IBM operates. Any reference to an IBM product, program or service is not intended to state or imply that only IBM's product, program, or service may be used. Any functionally equivalent product, program, or service that does not infringe any of IBM's intellectual property rights or other legally protectable rights may be used instead of the IBM product, program, or service. Evaluation and verification of operation in conjunction with other products, programs, or services, except those expressly designeated by IBM, are the user's responsibility.

IBM may have patents or pending patent applications covering subject matter in this document. The furnishing of this document does not give you any license to these patents. You can send license inquiries, in writing, to the IBM Director of Licensing, IBM Corporation, 500 Columbus Avenue, Thornwood NY 10594, U.S.A.

The following paragraph does not apply to the United Kingdom or any other country where such provisions are inconsistent with local law: INTERNATIONAL BUSINESS MACHINES CORPORATION PROVIDES THIS DOCUMENT "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow disclaimers of express or implied warranties in certain transactions; therefore, this statement may not apply to you.

This publication could include technical inaccuracies or typographical errors. Changes are periodically made to the information herein; these changes will be incorporated in new editions of the publication. IBM may make improvements and/or changes in the product(s) and/or the program(s) described in this paper at any time.

It is possible that this publication may contain reference to, or information about, IBM products (machines and programs), progamming, or services that are not announced in your country. Such references or information must not be construed to mean that IBM intends to announce such IBM products, prograaming, or services in your country.

Requests for technical information about IBM products should be made to your IBM authorized reseller or IBM marketing representative.

(C) Copyright International Business Machines Corporation 1996. All rights reserved.

Note to U.S. Government Users. Documentation and programs related to restricted rights - Use, duplication or disclosure is subject to restrictions set forth in GSA ADP Schedule Contract with IBM Corporation.

1.3 Software prerequisites 10

1.4 The software configuration menu 10

6.2.5 Settings for Windows applications 54

6.2.10 Tuning tips for games 58

6.3.5 Video settings 68

6.3.6 Mouse and touch sensitive settings 71

6.3.7 Settings affecting communications applications 72

IBM's goal is to provide a more responsive OS/2 that is simpler and more intuitive to use, and provides easy connectivity. To achieve that end, OS/2 Warp 4 addresses three major issues:

• Ease of use - changes to the Workplace Shell interface
• Voice Navigation and Dictation
• Ease of connecting to whatever networked environment you wish

OS/2 Warp 4 contains the following new features and functions:

• Voice Navigation and Dictation - Combining speaker independence and continuous speech, OS/2 Warp voice-enabled the operating system to navigate OS/2 and macros can be created to simplify repetitive tasks.
• Internet-Aware Desktop - New desktop objects for HTML, URL, Java class and FTP servers. These objects enable drag-and-drop from the Internet to the desktop and vice versa. The FTP object provides a folder view of a FTP directory.
• 256-color Exploitation - Visuals use more colors to give the user a 3D experience using texture, shadowing and curved edges.
• More System Fonts - A new condensed font style is added that uses less desktop real estate while improving legibility.
• File and Print Services Client - Converge LAN Requester and Peer for OS/2 to provide a single administration workstation. Users can share peer resources and administer peer and LAN resources. Objects on the desktop will inherit peer functionality in their context menus. The File and Print Client can connect to IBM-compatible networks such as:
• OS/2 Warp Server
• OS/2 LAN Server
• Windows NT Server
• PC LAN Program
• OS/2 Warp Connect
• Windows NT Workstation
• Windows 95
• Windows for Workgroups
• LANtastic for OS/2
• LANtastic for DOS
• PCOM Lite - Personal Communications/3270 (PC/3270) and Personal Communications/5250 (PC/5250)-- TCP/IP Entry Level 4.1 are replacements for TN3270, TN5250 and PMANT. The new emulators feature 32-bit code, full font set and automatic font sizing, custom color palette, improved cut and paste, APL support, command line (IND$FILE) file transfer, context sensitive help, popup keyboard, hot spots, PCMCIA and support problem determination aids. Users are limited to two sessions.
• Support for JAVA(tm) Applications - Full support for the Java language, Java runtime and Java applets.
• TrueType Engine - New support for TrueType, a ready source of inexpensive fonts.
• OpenDoc(tm) - Runtime support for cross-platform compound documents.
• Open32 Support - Support for subset of a Win32 APIs and messages to ease the porting of Windows applications.
• OpenGL - A high quality, portable 3D graphics API that will primarily enable technical and engineering applications (CAD/CAM) to be ported to OS/2 Warp.
• DPMI 1.0 Subset - Enablement for support of Win32s 1.25a compatibility and support of Borland 4 tools.
• Plug-and-Play (PnP) Support - Automatic detection and resource allocation for enabled devices as well as legacy ISA devices. An application is provided to view resource assignments. The framework has been established for more sophisticated operations in future releases. OS/2 Warp's PnP support does not include dynamic installation and configuration.
• Warm Plug/Warm Docking - Support for docking/undocking and the swapping of diskette drive and CDROM drive in specific models of IBM ThinkPad 755 series.
• Infra-red (IR) Support - ThinkPad-specific support for IR supporting the IrDA interface.
• Graphics Device Driver Model (GRADD) - New device driver model that reduces the IHV development effort and improves time-to-market.
• Device Driver Pak (DDPak) - A collection of OS/2 device drivers available from IBM and third-parties offered as-is for customer convenience.
• Display Data Channel (DDC) Support - DDC2 support for automatic enabled-display recognition and refresh rate setting.
• Realtime Musical Instrument Digital Interface (MIDI) - Framework and API for delivering high quality, realtime relative MIDI applications.
• Security Enabling Services (SES) - Enablement for operating system security services.
• System Dump Formatter - Utility to read and format system dump information.
• First Failure Support Technology (FFST) Probes - Capture all data relevant to the error at the time it occurs.
• DMI Interface - Support for the Desktop Management Interface, an industry standard way to manage Pcs.
• TME NetFinity - Software to allow administrators to view, monitor and initiate actions to ensure the smooth operation of PCs.
• Software Registration (ART) - Online registration tool to encourage customers to register the software electronically via a modem or the Internet, FAX, mail or even the telephone. Users are gently reminded periodically until registration is completed.
• Retrieve Software Updates - Automatic software retrieval and install of fixes and upgrades via the Internet.
• XPG4/ULS Support - Provide ability for easier application adaptation of international requirements such as native languages, local customs and coded character sets.
• Bonus Pak Changes - Collection of personal productivity applications including word processor, spreadsheet, charting, report writer, database, calendar, monthly planner, appointment book and phone book. IBM Works 3.0 contains fixes, memory management improvements, updated help, HTML support, updated word processor and graphics filters and a button bar for commonly used features.
• Remote Support for OS/2 - Allows IBM service representative to dial into your system to assist with problem determination and resolution.
• HyperACCESS Lite - Async communications program to access bulletin board services. The update includes maintenance and other refinements such as user-defined modem strings and more colorful icons.
• CompuServe Information Manager - Provides access to the CompuServe online service. CIM 2.03 includes fixes and removes the OS/2 registration requirement.
• HP JetAdmin and MarkVision for OS/2 delivers an advanced network printing solution that enables you to easily install, configure, query, and troubleshoot network-attached printers from your OS/2 Desktop.
• FaxWorks - Allows FAXes to be sent and received. Includes features of a basic answering machine. Supports drag-and-drop. FAX/data integration through a new interface for data applications to pass answered call to FaxWorks.
• AskPSP - AskPSP consists of an expert system with a natural language interface. It can be used to resolve customer problems on OS/2 Warp. This tool is also used by IBM Service and Support. The technical database, updated monthly, is available by subscription to the OS/2 Technical Connection CDROM.

The following features and functions have been updated in OS/2 Warp 4:

• Integrated install - Updated for OS/2 Warp, the install program gives the user an easy and an advanced path. The user can also selectively install, re-install or uninstall components.
• Controls and Visuals
• Visuals - Use more colors to give the user a 3D experience using texture, shadowing, and curved edges.
• Dialogs - File Open and Close dialogs provide tree views of directories and allow for filetype selection.
• Horizontal Style Notebook - Notebook settings add a horizontal tab option. This new style allows information to be presented more concisely without taking up much space.
• Close Button - Close applications with a single click.
• Folder Menu Bars - Menu bar access for common operations within folders. Users can also utilize the context menu (Mouse Button 2) for these same operations.
• User Interface Redesign
• Background Bitmaps - Bitmaps with texture and color schemes. Exploit use of more colors for higher degree of texture and depth.
• Arrange Options - Desktop and folder icon alignment according to user preferences.
• Pointers - Greater selection of pointers for user customization.
• Font Palette - System-defined or user-defined selection of fonts for text
• Color Palette - System-defined or user-defined selection of colors for visual controls.
• Scheme Palette - System-defined or user-defined selection of fonts and colors.
• Sound Schemes - System-defined or user-defined selection of sounds for system events.
• System Tutorials Update - Designed to help users get up-to-speed, with a focus on getting connected. Also includes a section on speech.
• Help System Updates - Enhancements to give the help author more control over content, presentation, and navigation.
• Migration Database - Database of optimal settings for popular OS/2, DOS and Windows applications. The database is used during install to migrate previously installed applications. Includes additional Windows and new Win32s applications.
• Async Read-Ahead - File system studies disk read patterns, anticipates a disk read and makes it available in memory to improve system throughout.
• Network Transports - NDIS device driver support for a variety of LAN adapters.
• Netware Client 2.11+ for OS/2 - This is the same level of code that was shipped in OS/2 Warp Connect 3.0 and OS/2 Warp Server 4 with fixes.
• TCP/IP
• DHCP/DDNS Client - Dynamic Host Configuration Protocol eases network administration by dynamically allocating and reusing IP addresses. Dynamic Domain Name Service simplifies network access, operation and change with dynamic resolution of IP addresses to IP hosts.
• Socks Security - Permits TCP/IP applications to access the Internet through many standard firewalls.
• FTP and TFTP Client and Server - File transfer to and from a remote host.
• Telnet Client and Server - Logon to and from a remote host. Client is now based on PC/3270.
• REXEC and RSH Client and Server - Execute command on and from a remote host.
• SNMP Agent and Manager - Communicate and obtain status information and manage network resources.
• WebExplorer - WebExplorer 1.1a supports HTML 3.0 and contains numerous fixes and updates.
• Remote Access Client - The Remote Access Client provides LAN access via dial connections (asynchronous, synchronous, ISDN or X.25). The Remote Access Client can dial into either a LAN Distance Connection Server or OS/2 Warp Server. In addition, a Remote Access Client can dial directly to another Remote Access Client to establish a virtual network.
• Mobile Office Services - Mobile Office Services transparently caches files while connected to an IBM-compatible or NetWare network and allows the user to continue using these files even when disconnected. When the network connection has been re-established, Mobile Office Services detects the differences between the cache and the files on the network and prompts the user for resolution.
• Win32s 1.25a Support - Win32s 1.25a application support.
• SOM/DSOM - Provides a language independent, cross platform architecture for sharing objects.
• Device Driver Update - Updates of many device drivers shipped in previous versions of OS/2.
• PCMCIA Enhancements - The latest level of code that supports IBM PCC and other OEMs. This consists of card and sockets services, cardbus and multi-function card support, an enhanced user interface as well as updates to warm docking.
• Advanced Power Management 1.1- Update to support suspend, resume and device management and control.
• Enhanced IDE Support - Enhanced to support the SMART standard for hardware failure alert and DMA capabilities for PCI IDE hard drives.
• Printer Support - New printer device drivers included.
• Multimedia Support - Direct Audio Routines (DART) used by speech navigation and dictation.
• DIVE Enhancements - Enhancements for hardware video accelerators, video capture, MPEG playback and full-screen support for applications needing high frame rates.

Some of the changes that gave OS/2 Warp 3 better performance than previous OS/2 versions are as follows.

• A new link parameter, /EXEPACK, which was used on many system files when OS/2 Warp was compiled, compresses resource and message files by 20 to 30%. These files load 2 to 3% faster, on average.
• Some page frames are "zero compressed" (similar to PKZIP) and put in areas of memory that is already allocated but that has extra space. This allows a page frame to be swapped from RAM to RAM (which happens in nanoseconds) versus swapping from RAM to the hard disk (which takes milliseconds). Up to 250 of these pages can be moved to different areas of RAM instead of the SWAPPER.DAT file.
• Many system DLLs (dynamic link libraries) are now being swapped out to the hard disk. This makes it much quicker to recall them. A lot of overhead is required to load a shared DLL from the file system versus saving it and its associated system data in the SWAPPER.DAT file.
• The SWAPPER.DAT file, when initialized, is put in the largest contiguous space of the FAT file system. Previously the system was not concerned with swapper fragmentation.
• The internal file system in the SWAPPER.DAT file now has a 4 KB cluster size instead of 2 KB. This change makes swapping a 4 KB page more efficient.
• Some system DLLs have been merged, so less overhead is needed to load them into memory and notify all tasks of their locations. For example, PMMERGE.DLL contains three old system DLLs: PMGRE.DLL, PMSHAPI.DLL, and PMWIN.DLL. Although these DLLs are still in OS/2 Warp, their function is simply to forward their calls to PMMERGE.DLL. Some other DLLs that are merged are DOSCALL1.DLL and MMPM.DLL.
• A method of addressing called basing puts some system DLLs at an absolute address to reduce the overhead of finding them in memory.

OS/2 Warp 4 contains all of the aforementioned changes as well as enhancements to the FAT file system (FAT deserialized and optimized asynch read ahead), CD-ROM asynch read ahead, page tuning many system DLLs, single input queue, as well as changes to the memory caching default size and system loader fixup scheme. Furthermore, there are changes made to improve TCP/IP throughput and faster response time for network administration commands.

OS/2 Warp 4 can be installed over the following systems.

• OS/2 1.3
• OS/2 2.0 and 2.1
• OS/2 for Windows
• OS/2 for Windows plus Service Pak
• DOS 3.1 or greater with Windows 3.1, 3.11
• DOS 3.1 or greater with Windows for Workgroups 3.10 or 3.11
• OS/2 Warp V3.0
• OS/2 Warp with WIN-OS/2 V3.0
• OS/2 Warp Connect V3.0
• OS/2 Warp Connect with WIN-OS/2 V3.0
• over itself
• in a totally empty partition

• OS/2 Warp 4 shipped at the equivalent of OS/2 Warp 3 with Fixpak 20 level.

Many OS/2 parameters can be modified during or after installation. This section overviews those paramters that can be modified during the installation process. Each of these tunable parameters will be discussed in greater detail later in this document. From the OS/2 Setup and Installation panel, there will be three menus: Options, Software Configuration, and Help.

• From the Options menu bar choice, you can:
• Install the selected features
• Format other drives
• Access an OS/2 command
• From the Software Configuration menu bar, you can specify the location for the swap file. You can also change the OS/2 and DOS configuration settings of your system.
• The OS/2 parameters that you can change are:
• Printer monitor buffer size
• Buffers
• Disk cache
• Maxwait
• Swap Minfree
• Threads
• Memman Protect
• Memman Swap
• Priority
• The DOS parameters that you can change are:
• Break
• Open FCBS
• Protected FCBS
• RMSize

The OS/2 operating systems, in particular OS/2 Warp 4, have evolved to take advantage of new technologies and strategies, such as client/server, voice navigation and dictation, internet aware desktop, JAVA support, etc. The resulting changes in the software infrastructure and system platforms require careful system planning to ensure OS/2 Warp 4 performs as a cost-effective information technology platform.

Personal computers have limited resources to handle increasingly complex operating systems and applications. Resources such as CPU, disk, and memory must perform well together and deliver adequate response times to satisfy a user's requirements. Hence, capacity planning becomes an important issue when determining the impact of OS/2 Warp 4 on your system. To ensure the correct level of resource to meet and support your computing needs, you need to:

By employing validated performance analysis and tuning methodologies and systems tools, you can accurately measure the utilization of your computer resources. This will the smooth integration of OS/2 Warp 4 into your existing environment.

Your system performance depends largely on how OS/2 Warp 4 and the underlying hardware work together. The 32-bit OS/2 Warp 4 operating system exploits the Intel 32-bit X-86 architecture through the flat memory model, native protected mode and virtual 8086 mode though paging and multiple virtual DOS machine sessions. The flat memory model allows for a very large (4 GB) single address space referred to as flat memory. Call/return times are reduced by eliminating the need to switch segments manifested in a typical 16-bit application.

While all previous 16-bit APIs are still completely supported, the remaining OS/2 Warp 4 APIs are now 32-bit, thus improving performance and enhancing function. In addition, OS/2 also provides support for advanced disk hardware, Direct Memory Access for its parallel port, and many industry standard audio and video devices. Like previous versions of OS/2, OS/2 Warp 4 supports all existing 16-bit and 32-bit OS/2 applications, most DOS applications including those that use EMS, XMS or DMPI memory in the full screen or windowed virtual DOS machine session. Windows application support is also included in OS/2 Warp by providing both Windows 3.1 and Win32s environments for running Windows applications seamlessly on the OS/2 workplace shell, in concurrence with DOS and OS/2 applications.

Several of OS/2 Warp 4's new features and enhanced functions require a powerful processor, more physical memory, and a larger hard drive. If you experience sluggish response time after installing OS/2 Warp 4 you can use some of the performance tools shipped with OS/2 Warp 4 to determine the cause of the problem and apply the following guidelines to obtain the best performance out of OS/2 Warp.

Without Voice Navigation and Dictation:

• Intel 486 DX 33MHz or higher processor
• 12MB to 16MB of RAM
• Installation by selecting options requires 100MB-300MB disk space
• Easy Installation (OS/2 Warp preselected options) requires 200MB disk space
• 640x480 resolution display with 256 colors recommended (SVGA monitor)
• IBM compatible mouse is required
• OS/2 compatible CDROM drive and 1.44MB 3.5" diskette drive "A"
• 14.4K or higher modem or network connection for Internet access
• OS/2 supported sound card for multimedia

With Voice Navigation and Dictation, OS/2 Warp 4 requires

• For speech navigation, Intel Pentium 75MHz or higher with 4MB additional memory
• For speech navigation and dictation, Intel Pentium 100MHz or higher with 8MB-12MB additional memory
• Installation by selecting options requires 100MB-300MB disk space
• Easy Installation (OS/2 Warp preselected options) requires 200MB disk space
• 640x480 resolution display with 256 colors
• IBM compatible mouse is required
• diskette drive A and CDROM driver
• 14.4K or higher modem or network connection for Internet access
• Sound card for multimedia or speech
• High quality microphone for speech (requires Active Noise Cancellation feature, ANC, for optimal performance.)

There are three areas where performance can be constrained: memory, disk, and processor.

• Memory constraints occur when you attempt to run more programs in memory than there is actual memory. The hardware allows OS/2 to execute beyond the real memory by allowing paging or swapping, i.e., to allow code and data to be moved from memory to disk. As memory becomes scarce, system performance degrades as access time to code or data written to disk now includes the disk access time required to read back into memory information that was moved out. It probably also required moving off to disk some information to make room for the returning information. This activity can cause extremely poor performance.
• Disk constraints occur when your application requires large number of disk accesses quickly. Since the hard disk must complete one I/O request before starting the next request, multiple requests can start to queue, causing poor performance. Between I/O requests, the hard disk must move the head from one location to the next. Thus a hard drive with slow seek time will become a major bottleneck if the support for high I/O rate is required.
• The processor constraints occur when processor-bound instructions and multiple concurrent tasks require heavy processor time. Since OS/2 is a true multitasking system, the likelihood of becoming CPU constrained is increased as the number of concurrent task execution increases.

The CPU plays a vital role in system performance. It is responsible for processing the millions of instructions necessary to produce work. The speed of the microprocessor (and also the amount of installed memory) typically has the largest effect on the performance of the computer system. The microprocessors system clock rate, measured in millions of clock steps per second, or Megahertz (MHz) dictates the speed at which the system runs. The faster the clock rate, the faster the performance.

Most CPU's in today computer systems are at least 80486 chips. The 80386SX, SL and SLC, as well as the 486SLCx have an external data path of 16 bits while 80386DX and 486DX and above processors have an external data path of 32 bits. This distinction becomes significant as the shift towards 32-bit software begins to proliferate the market. 32-bit instructions, processing on a 16-bit CPU must become two 16-bit instructions in order to be processed. The processor with a 16-bit external data path can result in about 10% lower performance than an identical processor of the same speed with a 32-bit data path. The following charts illustrate the various processor chips, their respective speeds and their cache option.

Feature 386SX 386SL 386SLC 386DX

Speeds available (MHz) 16 20 20 25 16 20 16 20 25 33

Internal Processor Cache N/A N/A 8K N/A

Internal Processing (bits) 32 32 32 32

External Data Path (bits) 16 16 16 32

Feature 486SLC 486SLC2 486SLC3 486SX 486SX2 486SL 486DX 486DX2 486DX4

Speeds available (MHz) Numbers represent both internal and external processor speeds. 25 40/20 50/25 60/20 66/33 75/25 100/33 20 25 33 50/25 25 33 25 33 50 50/25 66/33 75/25 100/33 100/50

Internal Processor Cache 16K 8K 8K 8K 8K 16K

Internal Processing (bits) 32 32 32 32 32 32

External Data Path (bits) 16 32 32 32 32 32

Note: The 486SLC2 processor runs internally at twice the speed of the rest of the system, such as 40/20MHz producing performance faster than a 25MHz 486DX, but less than a 50MHz 486DX. It is up to 271% better than a 20MHz 386SX, up to 99% faster than a 20MHz 386SLC, up to 53% faster than a 20MHz 486DX and up to 20% faster than a 25MHz 486DX. The 50/25MHz 486SLC2 is faster than an Intel 33MHz 486DX.

Feature Pentium P5 Desktop Pentium P54C Desktop Pentium P54LM Notebook

Speeds Available (MHz) Numbers represent both internal and external processor speeds. 60 66 75 90 100 120 133 150 166 200 75 90 100 120 133 150

Clock Multipliers (e.g., 2X = 133/66MHz) 1X 1.5X, 2X, 2.5X, 3X 1.5X, 2X

Internal Processor Cache 16K 16K 16K

Internal Processing (bits) 32 32 32

External Data Path (bits) 64 64 64

The Pentium Pro processor (also known as the P6 processor) is the latest generation processor family from Intel. The Pentium Pro processor includes significant architectural innovations and enhancements, like Dynamic Execution, multiple branch prediction capabilities, 256K L2 cache in the package, etc. The result is a significant boost in system performance -- especially with 32-bit software. Since OS/2 and especially OS/2 Warp 4 is a 32-bit operating system, the Pentium Pro processor delivers optimal performance with 32-bit OS/2 applications.

A summary of the performance data between P6 and P5 is given below. The experiment was done on a P6 processor running at 150 MHz, 256 KB Level 2 cache, 32MB of physical memory, 1.28 GB IDE hard drive, Matrox PCI display adapter card. All measurements were done with the display resolution set at 1024x768x256 except where it is noted. The P5 system is from Micron Technologies. It has a P5 processor running at 133 MHz, 256 KB Level 2 cache, 32 MB of physical memory, 1 GB SCSI hard drive, and Matrox PCI display adapter card. Again, all measurements were done at 1024x768x256 resolution except where it is noted. OS/2 Warp Connect was used on both systems with no special options or setup. OS/2 Warp was installed with multimedia support using the default installation options. The FAT file system was used in the study with default dynamic disk cache size.

There are five types of benchmarks used in the study:

In general, the average ratio of performance improvement for P6 over P5 is 1.63. In particular, the ColorWorks and DIVE benchmarks seem to gain the most, more than 2X improvement, while others tend to gain somewhere around 50%. The BRender 3-D library from Argonaut Tech which is built upon DIVE seems to take the full advantage of the P6 floating-point performance. Its ratio of performance is around 1.88, almost twice of that P5. It is also observed that applications that are computational intensive such as 3-D graphics rendering in the MicroStation benchmark gain about 1.5X on the P6 platform. The industry standard graphics test from GPC using the CDRS and DX viewsets also scores well on the P6, around 1.5 times over P5. The legacy OS/2 applications benchmark which is a mix of 32- and 16-bit code ran only 30% faster on the P6. The chart below illustrates the average measurements and the performance ratio between the two systems.

Processor: P6-150 P5-133 P6/P5

Manufacturer: Intel Micron Ratio

Memory Configuration: 32MB 32MB

Display Resolution: 1024x768x256 Improvement

Display Adapter: Matrox Matrox for P6

Applications Metrics

BYTEmark 2.0 iterations/sec 1.49

OPENGL - CDRS frames/sec 2.23 1.51 1.47

OPENGL - DX frames/sec 0.89 0.58 1.53

VISUAL AGE - Visual Builder seconds 80.4 104.4 1.3

MOVIE PLAYER - Bundy seconds 13.78 20.01 1.45

SAS seconds 16.4 27.84 1.7

Argonaut - Robot frames/sec 64 34 1.88

ColorWorks seconds 14.06 31.02 2.21

MicroStation seconds 31.55 46.94 1.49

Legacy Apps (Describe, Amipro, L123, IBM C++) seconds 61.5 80.5 1.31

DIVE frames/sec 273 128 2.13

Average 1.63

Geometric Mean 1.61

NOTES: 1. ColorWorks was run at display resolution 1024x768x16M

• Industry Standard Architecture (ISA) bus, often referred to as "AT bus". This bus is the 16-bit bus initially developed for IBM's AT (Advanced Technology) computers. The bus includes 8-bit slots for downward compatibility with earlier adapters, but includes 16-bit slots for improved, AT-compatible adapters such as 16-bit VGA adapters. It supports data rates of 8MB/sec.
• Micro Channel Architecture (MCA) bus. A proprietary 32-bit bus used in IBM PS/2 computers. This bus operates at 10MHz with burst data rates of up to 80MB/sec.
• Enhanced Industry Standard Architecture (EISA) bus. A 32-bit bus that, unlike the MCA bus, is backward compatible with ISA adapters.
• Peripheral Component Interconnect (PCI) bus. A 32-bit bus standard used in most new computers. This bus provides the greatest performance of all the above, operating at 33MHz with burst data rates of up to 132MB/sec.

The system bus plays an important role in balancing system performance. If you have a high speed CPU and video adapter on an ISA bus system, this system is not well balanced. The CPU is a great deal faster than the system bus. For example, if your processor is a 486DX 33MHz CPU and you have an ISA bus VGA video adapter (remember, ISA is 16-bit), the system has the capability to transfer data at speeds greater than 33MB/second but the ISA bus can only transfer data at 8.33MB/second. While this isn't a very practical example, it illustrates the potential for an imbalanced system caused by a system bus that is slower than the processor, particularly when running I/O intensive applications such as databases.

Memory, often referred to as random-access memory (RAM), is the computer's primary storage space used to execute instructions. Memory is generally available on the system board and/or on adapter cards. The speed of the memory is expressed in nanoseconds, with the lower the value the better.

Memory speed is important in that faster memory provides faster access to the information stored there. Memory location is also important. Memory on the system board (motherboard) can generally be accessed by the processor faster than memory located on an adapter card. The adapter card memory requires access through the system bus, slowing down the time from processor to memory.

Having adequate memory in a system will also reduce the disk access because paging or swapping is decreased. If the operating system does not have to handle paging I/O requests, system performance will improve. If the swap file is large, and changing from one application to another results in I/O requests, the system would benefit from additional memory or performance tuning.

It is important to note that the number of unique pages accessed during a sliding window of 12 snapshots (1 minute) taken on the same system but with 40MB of memory (unconstrainted memory scenario) is as follows.

OS/2 Warp 4 Connected 25.27MB

Fixed disk performance is important to the overall performance of a computer. The performance of a fixed disk refers to the rate at which information can be located and transferred between the fixed disk and the memory. The disks average seek time, average latency and data transfer rate determine how the disk subsystem will contribute to or hinder overall system performance. These disk performance statistics are generally available from the disk manufacturer.

Data on a fixed disk is stored in concentric rings, or tracks, on the disk surface. To read from a fixed disk, the actuator must first move the read/write head to the proper track. The average time it takes for the actuator to move the read/write head over the proper track is called the average seek time - usually expressed in milliseconds.

Once the read/write head is located over the right track, it must wait until the disk rotation brings the right part of the track under the read/write head. The average time it takes for this to happen is the average latency of the drive - also expressed in milliseconds.

Finally, after the proper track and proper part of the track are positioned under the read/write head, the information is transferred between the disk and the disk controller circuitry, one bit at a time in a continuous stream as the disk surface passes underneath the read/write head. The speed at which this is done is called the data transfer rate and is expressed in millions of bytes per second (MB/second).

It is important to consider the disk drives performance statistics when purchasing a computer system. These statistics play an key role in the overall performance of the system. The shorter the seek time and latency the better. The higher the data transfer rate the better. All of these factors determine how the disk subsystem will contribute to or hinder overall system performance.

OS/2 Warp 4 provides significant new function compared to OS/2 Warp Connect. Depending on the selections you make during installation, this version will require a maximum of 350 megabytes of hard disk space. The following table indicates how much disk space is consumed by each selectable component. It will be very useful in planning your installation.

• If you have a very large partition (> 1 GB) formatted for FAT, the install will require 550 MB due to the large cluster size required by the FAT file system.
• Failure to provide sufficient hard disk space will result in installation failures and will require re-installation.

During the installation process, you will be asked to select which of the following install options: Advanced Install or Easy Install.

If you selected the Advanced installation option then you would see the OS/2 Setup and Installation screen which enables you to configure the software. The list of software that you can select in OS/2 Warp 4 is as follows.

Disk Space Default

Assistance Center 10.03 MB

OS/2 Tutorial 4039 KB

OS/2 Command Reference 825 KB Yes

REXX Information 768 KB Yes

OS/2 Warp Guide User Interface Agent 4,367 KB Yes

Fonts 2.42 MB Yes

Courier 273 KB Yes

Helvetica 629 KB Yes

System Mono-Spaced 86 KB Yes

Times Roman 596 KB Yes

Courier (outline) 320 KB Yes

Times New Roman (outline) 259 KB Yes

Optional System Utilities 2.29 MB Yes

Backup Hard Disk 27 KB Yes

Change File Attributes 36 KB Yes

Display Directory Tree 33 KB Yes

Manage Partitions 233 KB Yes

Label Diskettes 33 KB Yes

Link Object Modules 450 KB NO

Picture Viewer 122 KB Yes

PMREXX 146 KB Yes

Recover Files 45 KB Yes

Restore Backed-up Files 35 KB Yes

Sort Filter 31 KB Yes

Installation Utilities 419 KB NO

Create Utility Diskettes 192 KB Yes

Service/Diagnostic Aids 542 KB Yes

Optional System Components 2.30 MB Yes

OpenDoc 5,842 KB NO

Voice Type 23,753 KB NO

Security 496 KB NO

Bonus Pak 36.42 MB NO

CompuServe 2,736 KB NO

HyperAccess Lite 656 KB NO

IBM Works 14,411 KB NO

Fax Works 1,266 KB NO

VideoIn 467 KB NO

AskPSP 4,008 KB NO

Remote Support Tool 1,441 KB NO

Printer Utilities

JetAdmin 560 KB NO

JetAdmin Port Driver 1,645 KB NO

MarkVision 4,879 KB NO

MarkNet Port Driver 5,238 KB NO

Tools and Games 24.19 MB

Enhanced Editor 1,933 KB Yes

Search and Scan Control 68 KB Yes

OpenGL 1.0 3D Library 4,905 KB NO

Optional Bitmaps 10,073KB NO

Solitaire-Klondike 2,762 KB Yes

Pulse 43 KB Yes

Chess 2,828 KB Yes

Mahjongg Solitaire 2,151 KB Yes

OS/2 DOS Support 1.54 MB

DOS Protected Mode Interface 29 KB N/A

Virtual Expanded Memory Management 19 KB

Virtual Extended Memory Support 9 KB N/A

WIN-OS/2 Support 6.11 MB

Readme Files 136 KB Yes

Accessories 1,039 KB Yes

Screen Savers 73 KB Yes

Sound 115 KB Yes

Multimedia Support 22.41 MB

Base Multimedia Support 19,294 KB Yes

Multimedia OpenDoc Support 2,929 KB NO

Software Motion Video 728 KB Yes

High Performance File System 470 KB Yes

Network Services

File and Print Client 13 MB

TCP/IP Services 28 MB

Remote Access Client 4 MB

System Management Client 7 MB

Netware Client 6 MB

Mobile Office Services 3 MB

• Selecting All Features During Advanced Install - A Full install requires approximately 275 megabytes for code and data, 25 megabytes for SWAPPER.DAT, and 50 megabytes for any additional programs and applications.
• Deselecting Features During Advanced Install - A single 350 megabyte partition is the least complex environment. You can deselect some features or install some features to partitions other than the OS/2 boot partition. The following features allow partition selection during install:
• Speech
• BonusPak
• Connect
• Java
• Multi-Media
• Windows Emulation
• Using Easy Install - Easy install selects a subset of OS/2 features for installation on the C: partition. This selection is based on the hardware configuration. For example, Multi-Media is installed if a sound device is detected. Some features that are not installed during Easy Install are:
• Opendoc
• Java Toolkit and Samples
• BonusPak
• Security
• Some optional BITMAPS
• Some documentation for Command Reference and REXX
• The selection panel for Connect allows installation of features consistent with the connection hardware and software. Easy install requires between 150 and 200 megabytes depending on your selections and hardware configuration.

Computers can provide video support on the system board or on an adapter. Video support is identified by the mode and resolutions that they can support. The greater the resolution and number of colors, the better the picture. Unfortunately, better resolution and more colors usually translate into slower system performance.

VGA (Video Graphics Array) was introduced in 1987 and soon became a popular video standard. VGA systems are capable of displaying 16 colors at a resolution of 640x480.

SVGA (Super VGA) refers to a group of high-resolution analog video adapters. SVGA provides higher resolutions and more colors than VGA adapters. Most SVGA adapters with 1MB of video memory are capable of displaying 256 colors in a resolution of 1024x768.

XGA (Extended Graphics Array) was introduced by IBM in 1990. It is an accelerated analog video adapter which uses a coprocessor to provide hardware-assisted drawing functions. XGA-2 is a newer version of XGA which includes performance improvements, higher refresh rates, non-interlaced mode and support for resolutions of 1360x1024 with 16 colors.

Many new video adapters are coming available that can't be classified as VGA, SVGA or XGA. They are using new and more powerful chipsets that show improved performance over SVGA by using on-board processors for some graphics functions. These chipsets are called accelerators. They include previously discussed XGA chipsets, as well as new accelerated SVGA chipsets such as S3, Weitek P9100, Mach 8/32 and Tseng W32i.

Two main features on your graphics card affect video performance—the first, the card's onboard memory, and the other the card's accelerator chip and bandwidth.

There are three common types of onboard memory found on graphics cards, Dynamic RAM (DRAM), Window RAM (WRAM), and Video RAM (VRAM). DRAM is cheap, but its single-ported design requires a system clock cycle to reset the chips before each refresh of screen data. VRAM is dual-ported, able to deliver data and reset in a single clock cycle for inherently faster performance. It's also more expensive than DRAM. WRAM is a form of VRAM that does faster fills and accellerates video playback and animation. WRAM is not only dual-ported, but also requires fewer transistors than VRAM, hence lowering costs while providing a few graphics-specific speed-ups such as support for aligned bit-block transfers (BitBlts).

The type of graphics memory you choose is almost as important as the amount. DRAM is the least expensive, but its single-ported design makes it relatively slow; dual-ported VRAM is more costly, but much quicker for true-color work. A variety of other single-ported (such as EDO DRAM and SDRAM) and dual-ported (such as WRAM) variations fall between the two on

the price/performance scale.

The graphics accelerator cards architecture, or more specifically, it's bandwidth also plays a crucial role in graphics subsystem performance. Bandwidth is defined as the amount of information that can move along the cards data path. The bandwidth between the graphics card's accelerator chip and its display memory can be as wide and quick as graphics card manufacturers care to make it. Over the last several years, 32-bit graphics accelerators have yielded largely to 64-bit accelerators, with 128 bit cards becoming more and more common. To keep refresh rates high, team a fast 64-bit or 128-bit accelerator with dual-ported display memory to handle the volume of data that true-color and multimedia applications demand.

Machines with a buffered UART (Universal Asynchronous Receiver/Transmitter), for example the 16550A will have better communications performance than machines without buffering, for example the 16450. The performance of DOS communications programs is particularly impacted by using systems with these buffered communications chips on their COM ports. The MODE command can be used to determine whether your system has this buffered UART. A buffered UART at level 16550A or equivalent is nearly essential for high speed communications. Use the command: MODE COMx, where x is the communications port number. If you see "BUFFER=N/A" then you do not have a buffered UART for that port.

HPFS has two limitations. Native DOS applications can't see HPFS formatted disks (although DOS programs running in OS/2 Warp can and there exists a driver to read HPFS disks from plain DOS), and the HPFS driver takes approximately 264KB of memory.

• Install only the file system needed by the operating system accessing your data. If you plan to boot a DOS or Windows system natively (via the dual boot option), then any data that will be accessed must exist on a FAT disk partition. However, If you will only be running DOS and Windows applications in an OS/2 Warp VDM (Virtual DOS Machine), then the file system can be either HPFS or FAT. Also, when accessing a file on a server, and the server file system is HPFS, you do not need to install HPFS on your local client machine. HPFS only needs to be installed on a computer when a partition on a local hard disk is formatted as HPFS.
• FAT is best suited for disk partitions that are 80 MB or less in size or that have a limited number of files installed. Usually, 256 files is a good target, with up to 500 acceptable. The number of files become important because FAT files are allocated based on a cluster size. The cluster size is determined by the size of the disk partition and can be 2K, 4K, 8K or higher. Since most file sizes are not an exact multiple of the cluster size, disk space is not optimally used. For example, installing DOS, Windows and OS/2 Warp on a 100MB partition resulted in 2.2 MB of disk space that could not be used. A 100MB partition will use a 2K cluster size. If you were to use a 540MB partition size, then your cluster size would be increased to 16K and a significantly greater amount of disk space will be lost.
• HPFS files are allocated based on a 512 byte granularity instead of a cluster size, therefore fragmentation is greatly reduced. Also HPFS is especially efficient when handling large partition sizes, > 100 MB, and large numbers of files, > 500. One thing you should look out for is to not allocate more than 5000 files in a sub-directory or directory. Allocating more than 5000 files can lead to degraded performance. The HPFS file system shipped with the OS/2 Warp product has a cache limit of 2 MB.
• It is recommended that HPFS be installed on systems with 16MB of memory or more and large disk partitions. If HPFS is not being used, you should remark the HPFS driver from the CONFIG.SYS file. This driver uses 264KB of RAM. If the HPFS statement configures a HPFS cache, the maximum amount of RAM consumed by the cache and the driver could be at much as 2312KB (264KB for the file system driver + 2048KB for the max cache size = 2312KB). Even if there is no HPFS partition on your system, it will cost between 200 and 250K in working set memory, as well as the space for the HPFS cache. If you are installing OS/2 Warp on an existing DOS and or Windows system, you should not install HPFS. When your system is up and running, you can check the working set of your system. If there is enough free memory and you wish to create a HPFS partition, then you can use selective install to install the HPFS support. Remember that any data stored in the HPFS partition can not be accessed if you boot your machine under DOS.
• Use HPFS if your application uses many small files or a very large data file like in database applications.
• Both FAT and HPFS file systems have disk caching, lazy writing and read-ahead. File system parameters can be changed after installation. The default values set by the installation procedure are good for average users.
• It is recommended that when you set up your hard disk, you create a minimum of 3 partitions. One will be for the operating system(s), one for your applications and static data files, and another for dynamic data files and temporary files. Decide whether you want to use Boot Manager or Dual Boot. If you select Dual Boot, then OS/2 must be installed with the FAT file system.

Response time is the key indicator used in measuring and tuning a system. It can be defined as the time interval from when a user initiates a process until that process has completed. For example, response time can be measured as the interval of time from when a user presses the enter key to initiate a database query until the data is displayed on the screen.

The productivity of a OS/2 Warp 4 user is largely dependent on adequate system response time. The user should be able to interact freely with the application without having to wait for the system to respond. Response time requirements may vary among users and even among applications. A three to fifteen second response time may be adequate for a complex inquiry application that is only run occasionally, but a sub-second response time may be required for more frequently used applications such as word processors or spreadsheets.

This section introduces the concept of performance monitoring, measurement, and tuning. It also covers performance tools shipped with OS/2 Warp 4 as well as tools available from external sources.

Monitoring the performance of critical system resources is invaluable in identifying the cause of performance problems. Performance monitoring is the first step to take in solving performance-related issues. There are numerous performance monitoring tools that measure response time, disk activity and other variables that impact the performance of the system. These tools will assist in understanding possible causes and subsequent resolution of the problem. Once the problem has been identified, the process for alleviating it can begin. This process, or tuning methodology, needs to be well thought out and documented through to resolution.

A benchmark is a test that measures the performance of a system or subsystem on a well-defined task or set of tasks. Test scenarios, or benchmarks, are designed to achieve consistent repeatable results during the tuning process. Benchmarks are necessary for creating an effective, systematic approach to performance tuning your system. The benchmark should represent a work environment very similar to the one being tuned. Benchmarking is essential for measuring progress while tuning a system.

Characteristics of a good benchmark include:

• Each test is repeatable.
• Each iteration of a test is started in the same system state.
• There are no functions or applications active in the system outside the ones being measured unless the scenario includes some amount of other activity going on in the system. Do not even have them started and sitting idle, since this will use up RAM resources and increase the likelihood of swapping.

Benchmarks can also be used as monitoring and diagnostic tools. By running a benchmark and comparing the results against a known configuration, you can potentially pinpoint the cause of poor performance. Similarly, a developer can run a benchmark after making a change that might impact performance to determine the extent of the impact.

The OS/2 WarpCenter provides a CPU monitor utility that shows the system activity and a disk space monitor program that shows the amount of disk space available in all partitions. In addition, there are many other operating system utilities that you can use to check your system integrity if you perceive a performance degradation has occurred. They are as follows.

• CHKDSK - disk integrity checking
• HDMON - hard disk monitoring
• PROFILER - file repairing
• PSTAT - process monitoring
• RMVIEW - resource allocation monitoring
• SYSLEVEL - operating system and corrective service level analyzing
• TRACE - events tracing

CHKDSK can detect lost clusters on your disk. These are parts of files that the system did not save completely and that take up space on your disk. If CHKDSK finds these, it prompts you with a message asking if you want to convert lost chains to files. If you type a Y (yes), CHKDSK converts these parts into files that you can examine and delete to save space on your disk. If you type an N (no), CHKDSK deletes these parts of files from your disk. The files CHKDSK creates from lost chains follow this naming convention: FILEnnnn.CHK (nnnn is a sequential number starting with 0000).

• To start chkdsk, type the following command in any OS/2 window.

• Type this command at a DOS command prompt to produce a memory storage report.
• CHKDSK gives accurate information only when a hard disk is not in use.
• CHKDSK does not work in DOS sessions on drives that have an ASSIGN, JOIN, or SUBST command in effect. Also, CHKDSK does not work on network drives.
• You should run CHKDSK occasionally on each disk to check for errors. If errors are found, CHKDSK displays the error messages and produces a status report. If you enter a file name after CHKDSK, the OS/2 operating system displays a status report that gives the number of noncontiguous areas occupied by the file. CHKDSK also produces a storage report.
• To search for and recover lost file clusters on the drive that is the hard disk from which you normally start the OS/2 operating system, follow these steps:

• To recover lost clusters on the drive that contains CHKDSK, you can also copy CHKDSK to another drive and run it from that drive by specifying the drive and path.
• If the /F parameter is not specified and there are open files, CHKDSK may report lost clusters on the disk. This happens when open files have been written to but the file allocation table (FAT) is not updated. If many clusters are reported as lost, use the /F parameter to search the disk.

The Hard Disk Drive Monitor uses Self-Monitoring, Analysis, and Reporting Technology (SMART) to monitor the status of physical drives in the system. A hard disk drive must be SMART-compatible for the Hard Disk Drive Monitor to be able to detect potential failures in that drive.

• To start the monitor, type the following command in an OS/2 window.

• The Hard Disk Drive Monitor program periodically checks the status of the SMART-compatible drives being monitored. In the Configuration screen, you can specify how frequently the status is checked. To view the Configuration screen, click on Options from the menu bar, and then click on Configuration. The value shown in the Configuration screen is the monitoring interval, in minutes. Click on the up-arrow or down-arrow button, or type a new value, to change the monitoring interval to any value from 1 to 60.
• The Hard Disk Drive Monitor checks status only if disk activity has occurred during the monitoring interval. The APM Scan option, which is enabled and disabled in the Configuration screen, allows the Hard Disk Drive Monitor to monitor drives while the computer is in an Advanced Power Management (APM) reduced-power state. If the APM Scan option is enabled, the Hard Disk Drive Monitor will check the drives every four hours, even if the computer is in a reduced-power state and no disk activity has occurred. If the APM Scan option is disabled, monitoring will occur only if disk activity has occurred during the monitoring interval. To enable the option, click on the radio button that reads, "Enabled."
• If the Hard Disk Drive Monitor indicates that a drive has a potential failure condition, you should back up that drive immediately, to avoid losing valuable software and data if the drive actually fails. However, the Hard Disk Drive Monitor is not to be used as a substitute for regular backups of the drive and may not detect all predictable failure conditions. Hard drives will also fail for reasons that are not predictable through use of SMART. In some situations, the Hard Disk Drive Monitor will not detect a potential failure condition soon enough to allow you to back up the drive before it fails.
• The main Hard Disk Drive Monitor screen displays a Drive icon for each hard disk drive being monitored. The colored indicator on each Drive icon indicates the status of that drive, as follows:
• Green indicator - OK

• Flashing red indicator - Alert

• Grayed-out indicator - Not SMART-compatible

Analyzes files on the specified drive and displays information about unusually fragmented files that do not meet the multimedia format requirements. You can use the PROFILER utility to correct files that do not meet multimedia requirements.

A fragmented file, although logically intact, is divided into blocks that are stored in different physical locations on the disk. Fragmentation is often necessary for the efficient use of disk space, but unusually fragmented files (those divided into small, scattered blocks) are not well suited for multimedia applications.

• To start the utility, type the following command in an OS/2 window.

• The PROFILER utility fixes a file that fails to meet the multimedia format requirements by copying it to a temporary file, deleting the original file, and then renaming the temporary file. This process reduces the amount of fragmentation in the file. The amount of fragmentation that is reduced depends upon the fragmentation of the drive's free space.
• If the 386 HPFS Local Security is installed, ADMIN privilege is required to execute the PROFILER utility.
• To capture all of the information displayed by the PROFILER utility, redirect its output to a file (for example, PROFILER /V D:\DIR > C:\PROFILER.OUT).

PSTAT displays process status information, such as current processes and threads, system semaphores, dynamic-link libraries, and shared memory. PSTAT helps you determine which threads are running in the system, along with their current status and current priorities.

• This command also aids you in determining why a given thread is blocked (waiting for a system event), or why the thread's performance is slow (low priority compared to other threads.) Moreover, it displays the process ID that has been assigned from each process. The process ID can then be used as input to the TRACE utility program for tracing on a per-process basis.
• To start PSTAT, type the following command in an OS/2 window

• Enter this command without a parameter to display information about the following:

This utility displays operating-system service level

• To start the program, type the following command in an OS/2 window.

• The following message will appear while SYSLEVEL checks the current corrective service level of your system

• Once the corrective service level has been determined the following will be displayed on your monitor.

• An example of the information displayed and an explanation of the displayed items follows:

This utility displays the allocation of the hardware resources in your computer.

• This is useful when resolving resource conflict or when installing a new piece of hardware on your computer.
• Type this command at an OS/2 prompt followed by any of its optional parameters. If no parameters are entered, RMVIEW defaults to the physical view which can also be specified with the /P parameter. This is useful when resolving resource conflict or when installing a new piece of hardware on your computer.
• RMVIEW Command: /P Parameter

• RMVIEW Command: /P1 Parameter

• RMVIEW Command: /D Parameter

• RMVIEW Command: /DA Parameter

• RMVIEW Command: /D1 Parameter

• RMVIEW Command: /DC Parameter

• RMVIEW Command: /DP Parameter

• RMVIEW Command: /L Parameter

• RMVIEW Command: /IRQ Parameter

• RMVIEW Command: /IO Parameter

• RMVIEW Command: /IOA Parameter

• RMVIEW Command: /DMA Parameter

• RMVIEW Command: /MEM Parameter

• RMVIEW Command: /HW Parameter

• RMVIEW Command: /SO Parameter

• RMVIEW Command: /R Parameter

The System Trace facility is used to record a sequence of system events, function calls, or data. The record is usually produced for debugging purposes. After the trace data is recorded, the System Trace Formatter is used to retrieve it from the system trace buffer and format the data to your display, printer, or file.

• This command is intended to be used with the assistance of your technical coordinator.

• The OS/2 operating system processes TRACE statements in the order in which they appear in the CONFIG.SYS file; the effects of the statements are cumulative. If any part of a statement is incorrect, the OS/2 operating system ignores the statement.
• If you do not specify TRACE in the CONFIG.SYS file, events are not traced. However, if you have a TRACEBUF statement in CONFIG.SYS, this allocates a trace buffer. Then, you can trace events by entering the TRACE command at the OS/2 command prompt.
• If TRACE=OFF or TRACE=ON appears in the CONFIG.SYS file without a TRACEBUF statement, the system allocates a 4KB trace buffer.
• If you do not specify TRACE or TRACEBUF statements in the CONFIG.SYS file, OS/2 does not allocate a trace buffer and system tracing is not available.
• If a system problem can be duplicated without a system failure, the TRACE OFF function allows tracing to be stopped after the problem has been re-created. This allows the state of the trace buffer to be preserved from the time the TRACE OFF command is processed.
• The tracing mechanism is performance critical; therefore, no statistical processing of recorded data is performed by the tracing routines.
• If you need to use the System Trace facility, your technical coordinator will provide the buffer size. When the trace is complete, you can use the trace formatter (TRACEFMT) to organize the data into a report. This helps you isolate causes of problems in the OS/2 system by formatting the information placed in the trace buffer by the Trace facility.
• An OS/2 enhancement to the Trace utility program allows you to trace a given process or set of processes, so that you can focus on the events of the specified process without intermixing events from other processes in the system. This reduces the possibility of trace-buffer overflow by minimizing the number of events which are recorded. Analyzing the formatted trace data is quicker and easier because only events of the specified process are recorded and displayed.
• You use the TRACEFMT utility program to format the information placed in the trace buffer by system trace. TRACEFMT is a Presentation Manager application running in a window. TRACEFMT can capture the current contents of the trace buffer, or can read an unformatted trace file. This trace entries can be searched, or a subset can be chosen to narrow the displayed entries to only those of interest. The TRACEFMT application provides choices on the menu bar. Your technical coordinator will analyze the formatted data to help diagnose your problem. You can use TRACEFMT as many times as required to diagnose a problem without having to restart the system.

TRACEGET is used to capture the contents of the trace buffer into a file. This file can then be loaded into the trace formatter, or sent to your service coordinator. A technical coordinator will analyze the formatted data to help you diagnose problems.

There are several useful external tools for monitoring and tuning the performance of an OS/2 system some of which are discussed in the following.

SPM/2 2.0 is an integrated package of powerful facilities that enable you to monitor resources such as CPU, RAM and disk on your local and remote OS/2 systems. SPM/2's ability to graph this resource information enables you to look at real-time data as well as saved data for any monitored workstation on the LAN. SPM/2 performs the following tasks:

• Collects critical resource utilization data from the CPU, memory, file I/O, swap file, FAT and HPFS cache, physical disk, printer, and communication ports.
• Records performance data to disk for processing at a later time.
• Provides a real-time or historical graphical representation of how system resources are being used (CPU, disk, RAM,and swap activity); it also has the ability to play back previously recorded data.
• Produces detailed resource utilization reports from recorded data that can be summarized by workstation, application, process or thread.
• Provides in-depth OS/2 memory analysis information that includes the working set and a view of OS/2 control blocks.
• Monitors remote OS/2 LAN Requesters and servers.

This tool, excellent for novice users, is a simple aid for tuning your system. It provides a way to change the CONFIG.SYS file and remove unnecessary files to free disk space.

CPU Monitor offers a selection of performance and analysis tools for OS/2 users. Using Presentation Manager graphics, CPU Monitor displays real-time information for estimated CPU utilization, OS/2 process relationships, and more. CPU Monitor enables you to dynamically suspend and resume execution for individual threads and helps you detect and stop runaway, invisible, and background programs.

This integrated group of applications is designed for real-time tracking and performing capacity planning functions for system resources including CPU, disk, memory, and applications. OSRM/2 monitors most of the paramters that SPM/2 does but it also works on symmetric multiprocessor (SMP) systems.

In general, performance tuning consists 4 steps:

A sound understanding of the problem is critical in monitoring and tuning the system. You should understand the problem in detail rather than at a high level like "the system is slow". Ask specific, pointed questions, like:

• If the system is slow, how slow is it?
• What activity is slow?
• How long does the activity usually take?
• Does the problem occur when using a particular application, when accessing a disk, or communicating remotely?
• What performance do you expect or believe is reasonable?

Be able to quantify the problem and when possible, obtain supporting data. As you are obtaining additional data, your problem description should become more focused.

You should also find out all you can about what has recently been done to this system that may have affected its performance. You may find that additional software has been installed, additional communications sessions were defined, or even that some of the memory has been removed.

Also consider the environment this system is working in. Environmental factors that can affect performance might include:

• Stand-alone or LAN environment
• Communication links
• Hardware configuration (memory, disk space, CPU, coprocessor)
• Applications running concurrently

Once you understand the problem, you should define a realistic goal for improvement. How will you know when the performance is acceptable? It may be easy to identify your desired goal, but you might not immediately know whether this goal is realistic.

After you have identified the problem and the environmental factors that affect it, you can narrow down possible causes and solutions. This involves identifying possible bottlenecks, verifying whether they are indeed bottlenecks, and devising possible solutions to alleviate them. Be aware that once a bottleneck is identified and steps are taken to relieve it, another bottleneck may suddenly appear. This may be caused by several variables in the system running near capacity.

Bottlenecks occur at points in the system where requests are arriving faster than they can be handled, or where resources, such as buffers, are insufficient to hold adequate amounts of data. Finding a bottleneck is essentially a step-by-step process of narrowing down the causes of the problem:

• Identify the hardware component
• In a networked environment, try to determine whether the bottleneck is occurring on the client or the server. If neither system seems to have a problem, check the physical network and all connectivity devices.
• Identify the resource that is affected
• A performance monitoring tool can be used to monitor system resources and indicate which resource is being overused.
• Consider what is happening internally when the function is performing poorly
• Consider all the steps undertaken by the poorly performing function. For example, if the system is sorting a remote database file, there should be heavy CPU and memory utilization. If you see a lot of disk activity this may indicate excessive memory paging and/or poor cache utilization.

Now that you have identified potential bottlenecks and have selected a performance monitoring tool, you need to have a plan to move forward into the tuning process. There are five basic components in a performance tuning methodology that apply to all cases:

• Define the problem
• Hypothesize solutions
• Design tests
• Set values and run tests
• Analyze the results

Use the process described earlier in this chapter to ensure a complete and thorough understanding of the problem.

Narrowing in on the cause of a performance problem is an iterative process, involving identification of possible bottlenecks and verification of whether they are indeed bottlenecks. It is important to make sure the correct bottleneck has been identified before pouring in resources to alleviate it. If not, unnecessary expense may be incurred by adding resources that do nothing to improve performance. For example:

• RAM could be added to a database server in order to increase the size of the buffer pool, only to find out that the problem is inappropriate indices, not a small buffer pool.
• A faster CPU could be purchased, hoping to improve response time. Although response time may indeed improve slightly with a faster CPU, it might later be discovered that the real bottleneck is insufficient buffer space on the adapter card or a slow fixed disk (in the case of an I/O intensive application). Hence, it is important to think through several possibilities and test them out. Once a bottleneck is identified, and resources have been applied to relieve it, another bottleneck may suddenly surface. This may be because several things in your system are running near capacity. Note however that there will always be something in the system that is the gating performance factor. Depending on the workload, this factor may or may not be an actual bottleneck.

To increase the capacity at a bottleneck, think of ways to offload the work arriving there or improve its capacity, through hardware or software. From there, come up with hypotheses of things that may relieve bottlenecks and improve performance. Possible examples follow, though in no particular order. Note that changing the values of system parameters may only be one avenue to pursue.

• Faster hardware - e.g. CPU, DASD, communications medium (i.e. adapter or connection)
• Parameter tuning
• Additional RAM or disk space
• Changing the topology - i.e. how many workstations are going through (or to) what servers, controllers, etc.
• Using or increasing a memory cache
• In a swapping system, perhaps reducing the size of buffers or numbers of active applications.
• Application redesign

Given the possible solutions, design tests to try them out. Make sure that these tests are repeatable. Approach the testing effort methodically, testing each hypothesis one at a time.

• Prioritize your ideas from "most likely to have an impact with minimal effort" to "least likely to have an impact with great effort".
• Start with a single hypothesis and decide what you are going to do to prove whether it is true or false. For example, if your hypothesis is to "add more buffer space," then only change the parameter needed to accomplish this.
• If the hypothesis is proven false, restore the parameter and move to the next hypothesis and start the cycle again.

Setting new variable values and running tests is probably the most iterative part of the entire process. In this stage, response time should be measured for specific, repeatable events, both before and after changing a value.

• Change only ONE thing at a time. Changing more than one variable will cloud results, since it will be difficult to determine which variable has had what effect on system performance. The general rule may perhaps be better stated as "change the minimum number of related things." In some situations, changing "one thing at a time" may mean changing multiple parameters, since changes to the parameter of interest may require changes to related parameters. (For example, in Database Manager, changing the BUFFPAGE parameter may require an increase to the SQLENSEG parameter.)
• Start in the same state. Start each iteration of your test with your system in the same state. For example, if you are doing database benchmarking, make sure that the values in the database are reset to the same setting each time the test is run.
• Check for Control. You can never control your testing too much. This mainly involves carefully documenting each step in your tests - to the point of being scientific. This will ensure that you're not doing things differently from run to run, and that you're not introducing new variables that could cloud the results. Writing things down is also essential for going back and recalling exactly what you have and have not tested since it is easy to forget!
• Work on one problem at a time. It is easy to lose control and focus when working with a large number of parameters and components. Additionally, it is likely that anomalies or problems will come up during testing, such as functional problems. While it will be necessary to resolve problems that obstruct your goal of identifying bottlenecks and improving performance, problems that are not directly related to this may need to be put aside for a while, or given to another group to work on in parallel. The point here is to maintain your focus.

Analyzing the results of your tests involves being able to interpret and explain the data you obtain in order to understand what's going on in the system. Information explaining the results should be available with the software tools used for the testing. SPM/2 has very detailed help screens that explain all the parameters it monitors.

After installation, there are changes that you can make to improve the performance of OS/2 Warp 4. Some of these changes like changing desktop settings are general in nature and will make slight improvements in the overall system performance. These changes can be done by any user and are listed in this section. Other changes are more specific for certain system components such as the system swap file, file systems, the CONFIG.SYS file, etc. These changes require certain knowledge of how various OS/2 components interact with each other. Such changes are listed in the next section, "Advanced OS/2 Tuning."

The OS/2 WarpCenter is a customizable, object-based status bar that can remain on top of all maximized windows. You can use the OS/2 WarpCenter to perform OS/2 desktop operations and display information about your system. You can display battery usage, system activity, and disk space monitors. OS/2 WarpCenter monitors the CPU activity on your computer and displays it through the system activity pulse. It indicates periods of low, medium, and high activity.

• The system activity monitor program runs at a very low priority level, therefore, it does compete with your application for CPU cycles. For optimal system performance, you can turn the monitor program off.
• To turn the system activity monitor program off:

OS/2 Warp 4 provides the OS/2 WarpGuide folder which contains guidance objects for selected computer tasks. Some objects have cue cards available to assist you with each step of the task.

• OS/2 WarpGuide is a task-mentor who uses cue cards or wizards to help you complete a task. As a result, anytime you start a task that OS/2 WarpGuide knows about, a OS/2 WarpGuide button appears on the title bar of the window you are using. If your profile indicates you are a novice for the current task, OS/2 WarpGuide automatically shows you cue cards and shades unneeded parts of the window. The shading does not change the way a window works. You can still click on the areas under the shading. If your skill level is intermediate, OS/2 WarpGuide shows only cue cards. If you are an expert, cue cards are not shown, but you can access a cue card for the current task by clicking on the OS/2 WarpGuide button.

• The appearance of the cue cards will affect your system performance because it requires processing time. Therefore, you should turn the cue cards off if you feel you don't need the OS/2 WarpGuide assistance.
• To permanently turn cue cards off:

Animation is the process of drawing boxes on the screen that appear to grow in size culminating in an open folder or session. This gives a nice appearance when drawing the screen objects. Disabling animation will save a small amount of system resource that would otherwise be used to expand and contract graphical brackets around folders as they are opened or closed

• On memory constrained systems, the performance when opening folders and starting sessions can be improved by disabling animation.
• To disable desktop animation:

During installation, OS/2 Warp detects your system display adapter hardware and installs the appropriate display drivers and resolutions. However, the higher the resolution, the more memory that is used. Very high resolution and color support can require 100 to 200K of physical memory. Since the display drivers and resolutions have great impact on your system performance, you can tune your display subsystem as follows.

• To change the display resolution:

• To change the display driver:

• If you are changing from one type of SVGA hardware to another type, it is best if you change the display driver to VGA before you change hardware, then change back again to the correct display driver type. This will keep you from having a display/hardware mismatch that could necessitate reinstalling OS/2. You can also revert to VGA during system boot.

The OS/2 system responsiveness may be better if you minimize frequently used applications between use since accessing a minimized application is always faster than starting the same application. This is also true when frequently working with the same folders.

• When you have applications or folders that you use often, leaving them open at shutdown (or placing them in the startup folder) will cause them to be loaded at system boot time, thus allowing you to access them much faster. However, placing the application in the startup folder will prolong the system boot time because it has to start the application.
• You can use the Minimized Window Viewer to view the applications and folders that have been minimized.

Performance when starting applications can be improved by reducing the application load time.

• To reduce the search time for OS/2 to locate the application, start the application from its own directory or call the application with a fully qualified path.
• Using the icon to start the application is optimal because the icon will have the exact path for the executable file if properly installed.

Applications and folders can be opened several ways. You can open your application faster if you know how to take advantage of the Startup Folder and the OS/2 multitasking feature.

• Applications that were left open at shutdown will reopen at boot time after the Workplace Shell is started.
• Applications placed in a STARTUP.CMD file will also be processed at boot time, after the Workplace Shell is started.
• Applications and folders placed in the Startup Folder will be opened with the processing overlapping the startup of the Workplace Shell.

• Run only programs that are necessary. Close all applications that are not needed.
• Verify that the settings are correct for the application, especially DOS applications that may poll, so that useful tasks are always processing.
• Limit background processing to improve foreground performance by increasing or reducing the application's session priority.

You should use solid colors and avoid the use of bitmaps for desktop and folder backgrounds. These particular options use more memory and require more processing time to display them.

Deselect the System Sounds options, unless you like the sounds when opening and closing your folders. It costs between 250 and 300K in working set just to hear the sound. An additional 40K or so of working set can be saved by executing DINSTSND.CMD in an OS/2 command session. This will unhook the system sounds from the OS/2 desktop. To get them back, you would execute INSTSND.CMD.

There are many fonts supplied with OS/2 and even more supplied with various word processing and graphics applications. Avoid the temptation of installing more fonts than are necessary by installing only the ones that you will use. You can always install others later if you should need them. This simple exercise of restraint will improve performance when loading applications that use fonts. Each installed font uses at least 2KB of memory, even if it is not being used. This number increases significantly when you actually use the font in an application.

When you have a choice of choosing which kind of fonts to install on your system, remember that outline fonts are more efficient than bitmapped fonts because the characters are cached into memory. With bitmapped fonts, the entire character rendering is loaded into memory. Bitmapped fonts are defined for a specific screen resolution and display type, whereas outline fonts are scaleable and tailored to the specific device installed on the system.

Mouse pointers are basically bitmaps. The amount of memory used will be affected by which mouse pointer style you choose. If you activate the comet cursor, this will cost additional memory and processing time whenever the mouse is being used.

It is important to close folders and applications when they are no longer being used. Although a large portion of unused programs will be swapped out, some portion of every program is non-swappable and will use memory for as long as the program is active. The more active programs in the system, the more memory will be used.

As placed ( or Flowed) icon views offer faster performance than Gridded views for folders containing multiple objects. The operating system has to retrieve an objects coordinates from the OS2.INI if a Gridded view is being used. Flowed views display one object after another without keeping track of specific coordinates.

You should also restrict the use of programs which, as a class are more decorative than useful. That is, the amount of function they provide is low compared to the memory they use. These programs, like screen savers, clocks and animated icons should not be used on systems with a limited amount of memory.

There are many more tuning parameters for an OS/2 system than discussed in the previous section. This section will detail advanced tuning parameters and explain how they can be used to further improve your system performance. This chapters focus will be on tuning parameters that optimize the Workplace Shell, task scheduling, CONFIG.SYS and AUTOEXEC.BAT files.

The Workplace Shell starts automatically when OS/2 is booted. The Workplace process is the one under which all the Workplace Shell classes are loaded and initialized. Therefore, objects representing Workplace Shell classes and their subclasses must run on this process. The Workplace process is actually launched from the Shell process, which is the process indicated in the SET PROTSHELL= statement in the CONFIG.SYS file. Once the Shell process is running, it starts the Workplace process. The Shell process is responsible for restarting the Workplace process in the event that it is ended as a result of a trap.

• The PROTSHELL= statement in the CONFIG.SYS file indicates which process is to be launched as the Shell process.
• The SET RUNWORKPLACE= statement in the CONFIG.SYS file indicates which process is to be the Workplace process.
• In the default configuration, both the PROTSHELL and RUNWORKPLACE environment variables are set to PMSHELL.EXE.
• PMSHELL.EXE is designed to distinguish between being started as the Shell process versus being started as the Workplace process.

Preemptive means that the multitasking activity needs no cooperation from the executing programs. The operating system maintains control over executing programs, and stops, or preempts, them when their time slice with the CPU is over or when a higher priority program is ready to run.

CPU scheduling is based on the following four priority classes, ranked in order from lowest priority to highest:

• Idle-time (priority 1)
• Regular (priority 2)
• Server (priority 4)
• Time-critical (priority 3)

Each class has 32 levels of execution ordering. Scheduling parameters are user-selectable at the time the system is started or can be varied dynamically based on system load.

Depending on a thread's priority class and level, the operating system periodically gives each thread in each process a small slice of CPU time. Threads with higher priorities always run before threads having lower priorities. A thread runs until its time is up or until a thread with a higher priority is ready to run. At that time, the operating system preempts the thread and starts another thread. Threads can also voluntarily relinquish the CPU (for example, by calling DosSleep).

The amount of time in each time slice is defined by the TIMESLICE command in the CONFIG.SYS file. The TIMESLICE command can be used to customize the size of the time slices that a thread gets. The default is for the operating system to dynamically vary the size of the time slice based on the activity of the thread and the overall system load.

When a thread is created (using DosCreateThread), it inherits the priority of the thread that started it. DosSetPriority enables threads to change their priority classes and levels in response to changes in their execution environments. DosSetPriority enables a thread to change its own priority, or the priority of any thread within its process. DosSetPriority also enables changing priorities for the entire process and for descendant processes. Within each class, the priority level of a thread can vary because of a DosSetPriorty request or, if dynamic priority variation is being used, because of action taken by the operating system.

• The TIMESLICE command in CONFIG.SYS sets the minimum and maximum amount of processor time allocated to processes and programs for both OS/2 and DOS sessions.
• The first parameter selects the minimum TIMESLICE value in milliseconds. This value is the minimum amount of time a thread is to be processed before yielding the processor to a thread of the same priority level. This value must be an integer greater than or equal to 32.
• The second parameter selects the maximum TIMESLICE value in milliseconds. The value of this parameter is the maximum amount of time a thread can be processed before yielding processor time. This value must be an integer greater than or equal to the minimum value, and less than 65536.
• The default is dynamic time slicing, which varies the size of the time slice depending on system load and paging activity. Dynamic time slicing is designed to give the best performance in most situations.

OS/2 provides a considerable amount of flexibility in the settings in the CONFIG.SYS file. You can change these settings for exploration and