Integrated SCSI Driver: AIC7870.ADD

The integrated SCSI controller on the PC Server 330 system board requires an updated AIC7870.ADD device driver. The old AIC7870.ADD that comes on the OS/2 Diskette 1 will not recognize the controller. An updated AIC7870.ADD must be copied on the OS/2 Diskette 1. The correct AIC7870.ADD will have a date stamp of 1996 or later.

To get a current AIC7870.ADD, access the Adaptec Web Pages (http://www.adaptec.com) and go to the Service and support area, OS/2, and download 7800OS2.EXE, or, via FTP, ftp.adaptec.com/pub/BBS/os2/7800OS2.EXE 7800OS2.EXE is a self-extracting compressed file, so just execute it to obtain it's component parts, including AIC7870.ADD.

Copy this new AIC7870.ADD to the OS/2 Diskette 1. If there is insufficient space to copy the new driver, you can make room by erasing the AIC7770.ADD file and other AIC*.ADD

Add two additional lines to the CONFIG.SYS on Diskette 1:

  BASEDEV=AIC7870.ADD
  SET COPYFROMFLOPPY=1

Completing OS/2 Install

• Support for desired printer(s): You should always define at lea st one printer type / queue, even if you do not have a printer physically attached.
• MultiMedia support: You probably will not have any 'sound cards' or other Multimedia devices, so this will be 'none'.
• VGA Display adpter: Just accept the basic standard SVGA support. You will need to install the SVGA upgraded drivers later.
• CD-ROM Device Support: Should default to "IBM CD-ROM II, Enhanced CD-ROM". This is the correct setting.
• SCSI Controller: Make sure this says 'none', otherwise the wrong level of the AIC7870 driver will be loaded from the CDROM. (Having the second instance of 'BASEDEV=AIC7870.ADD' will insure the correct statements are placed in the target CONFIG.SYS.)

In Summary

1. Make room on the Diskette 1
2. Copy a recent version of AIC7870.ADD to the diskette
3. Copy IPSRAID.ADD to the diskette
4. Add the following statements to the top of CONFIG.SYS on Diskette 1

      BASEDEV=IPSRAID.ADD
      BASEDEV=AIC7870.ADD
      SET COPYFROMFLOPPY=1
   

5. Specify None to SCSI Adapter Support on the "System Configuration" screen.

Material Available Electronically There are a lot of information files, device drivers, and 'fixes' available electronically from several sources. This section will list many of these files and the sources where they are available.

OS/2 and PC Related Files

IBM PC Company BBS

FTP Site

PC Company Web Page

To find files related to the PC Server 330 and PC Server S/390, use the Searc h function, searching on:

Server 330

To find all files related to the PC Server 330

ServeRAID

To find all files related to the ServeRAID adapter

Server 390

To find all files related to the S/390 component of the PC Server S/390

PC Company Configurator

P/390 And S/390 Related files

Temporary fixes to correct specific problems may sometimes be made available. Users should only install these fixes under the specific direction of IBM Support, or P/390 developers. Such fixes may be available via diskette from IBM Support or via FTP from: ftp://P390.IBM.COM

      login in as user anonymous,

       password=your_net_id       

      review the  'read.me' file

PC Server S/390 Home Page

Online Discussions

There is a "PC500390 CFORUM" available on IBM's "TALKLINK / DIAL IBM". (IBM employees can access this CFORUM on either IBMPC or IBMVM conference disks.)

There is an Internet P/370 - P/390 Discussion group:
P370-L @ PUCC.PRINCETON.EDU
Send 'SUBSCRIBE' request to listserve@pucc.princeton.edu

Large Partitions - CHKDSK Problems There is a known limitation in the current OS/2 CHKDSK program provided with Warp Server. This problem may appear in either of two ways:

1. CHKDSK takes a very long time to complete (45 - 60 minutes is not unusual).

2. CHKDSK fails or never completes

The problem is caused by the current CHKDSK only using memory below 16 megabytes. Since CHKDSK uses a form of in-memory 'bit-map' of the partition information, large partitions can effectively exceed this 16MB limitation, causing CHKDSK to fail. In general, our experience indicates the following results for different partition sizes:

01 - 10GB

Generally no problem, although 8-10GB partitions may take up to 15 - 20 minutes to CHKDSK.

10 - 14GB

Will usually complete successfully, but may take up to 60 minutes.

> 14GB

Will frequently fail outright, or never complete.

Recommendations:

Do not AUTOCHECK large partitions:

CHKDSK  D:  /C /F
CHKDSK  E:  /C /F
CHKDSK  F:  /C /F

Avoid Large Partitions:

Other Comments

A 'FixPack' for CHKDSK

PCI Slot Usage for P/390

The initial PCI bus design specified a maximum of three PCI slots. As use of the PCI bus grew, it quickly became apparent that three slots would not be sufficient for many systems, so a solution was devised. In effect, one of the three slots is used as a 'bridge' to another PCI bus, normally referred to as the 'Secondary PCI Bus'. This secondary bus is functionally equivalent to the primary PCI bus in most regards. There are, however, some considerations caused by this "bridged bus" design:

• Adapters on the primary bus will experience slightly better performance / throughput than those on the secondary bus. Therefore, adapters which cause large amounts of data transfer should use the primary bus. For example, the ServeRAID adapter should always be on the primary bus since it normally experiences the highest data transfer rates in the system.
• Due to timing characteristics, the PCI P/390 adapter must be installed in one of the secondary PCI bus slots (slots, 3, 4, 5 on the ESx or 4, 5, 6 on the PB0).
• PCI Adapters that frequently transfer data between them should be placed on the same bus (primary or secondary). For example, the PCI P/390 adapter and the S/390 Parallel Channel Adapter should both be plac ed on the secondary PCI bus. (Note that, since the ServeRAID adapter should always be placed in slot 1 for performance reasons, the only way that the PCI P/390 adapter and one or two S/390 Parallel Channel Adapters could be placed on the sa me bus is to put them on the secondary bus.)

Note that the slot numbers (1 to 9) are numbered from the bottom of the system upward. Slot 1 is at the bottom of the system and slot 9 (ESx models) or slot 6 (PB0 model) are the top adapter slots.

Checking and Resetting PCI Interrupts< /H2> (Updated 11/17/1997)
In some configurations, the PC Server 330 will assign the same interrupt level to multiple PCI adapters. Many PCI adapters can use these shared interrupts with no problems, but attempting to use a shared interrupt on the P390 adapter may cause a problem. The specific symptom this sharing of an interrupt will cause is termed a "Pink-Hang": the system comes to a complete stop (hangs) and the P390 activity window is all "Pink". This symptom is intermittant and may not appear for several days at a time. If you have (or suspect you might have) this problem, follow the procedure below to check for shared interrupts and to reassign the interrupts to eliminate such sharing.

One alternative to this process (below) will is to apply the new fixpack level 2.3.3. P390 code at Fixpack 2.3.3 should eliminate this shared interrupt exposure. (Fixpack 2.3.3 is functionally equivalent to P390 code level 2.4.) Even with the Fixpack or 2.4 version of P390 code, we recommend that the P/390 adapers not share interrupts unless absolutely necessary.

Checking For Shared Interrupts

1. Shutdown and power-off the system.

2. Remove the PC Server 330 cover and note what adapter cards are in what slots, especially the P390 adapter. (The P390 adapter will usually be in slot-4; the forth one from the bottom.) Replace the cover.

3. From a completely powered off condition, power on the PC Server 330 As i t is going through the Power On Self Test (POST), press the F1 key to cause it to enter the Configuration and Setup Utility.

4. From the main menu of the Configuration and Setup Utility:

      Select System Information    then
      Select   PCI Routing        
   

   You should now see a list of the integrated (planer) PCI adapters and adapters in PCI slots. For each adapter, there will be a notation of is routed to IRQn.

5. Check the "routing" for the slot where the P390 adapter is located (again, usually slot 4).

6. See if there are any other adapters/slots that are routed to this same interrupt (IRQ).
   • If any other adapters/slots are routed to this same interrupt, t hen you need to change the default interrupt settings to halt the sharing of the P390 interrupt.
   • If no other adapters/slots are routed to (share) this same interrupt then this system is not subject to the "Shared Interrupt Pink Hang". You can restart the system and resume normal operation.

Changing the PCI Interrupt Assignments< /A>

Table 1. Interrupt Summary Table

Interrupt #	Type (EISA/PCI)	Assigned use
0	Fixed	System Timer
1	Fixed	Keyboard
2	Fixed	Redirected to high INTs
3	Fixed	COM2
4	Fixed	COM1
5
6	Fixed	Floppy Disk
7	Fixed	LPT1 (printer)
8	Fixed	Realtime Clock
9
10
11
12	Fixed	Mouse
13	Fixed	Coprocessor
14
15

The Interrupts marked as Fixed are standardized and pre-assigned interrupts. These should not be changed unless you understand exactly and completely what you are doing. (The COM2 interrupt is the one which could most easily and reasonably be reassigned to something else.)

Your first step is to fill in the blank entries in the table based on your system. Most or all of the interrupts can be determined from the Configuration and Setup Utility. However, on ESx models of the PC Server 330, you may need to use the "EISA Configuration Diskette" to determine interrupts of any installed EISA adapters. To do this (for ESx models only):

1. From a powered-off mode, place the "EISA Configuration Diskette" in the diskette drive and power the system on.

2. When the system boots the diskette, reply 'Y' (yes) to continue into the configuration utility.

3. From the Primary Menu, select "Step 3: View or edit details" .

4. From the "View or Edit details" screen, press F7 for "Advanced"

5. From the "Advanced Menu" popup, select: "View Additional System Information Menu"

6. From the "View Additional System Information Menu" popup, select "Used Resources"

7. The first screen of the "Used Resources" will be a list of all of the Interrupts which could possibly be reassigned.

8. Now, use this information to complete the table above.

At this point, you will find one of three possible conditions:

1. There is an interrupt unused and available

2. There are no unused interrupts, but a embedded/integrated device or a 'Fixed' device can be disabled and the interrupt reassigned.

3. Not enough interrupts can be freed up by the above process, and some sharing of interrups between PCI adapters is required.

A Interrupt is Available:

Can reassign a fixed or integrated Interrupt:

Must share Interrupts on other PCI devices:

Reassigning the P390 Adapter Interrupt

1. From a powered off condition, power on the PC Server 330 As it is going through the Power On Self Test (POST), press the F1 key to cause it to enter the Configuration and Setup Utility.

2. From the main menu of the Configuration and Setup Utility:

      Select Advanced Setup        then
      Select   PCI Bus Control         then
      Select     PCI Interrupt Requests    then
      Select       PCI Bus Control    
   

   You should now see a list of the PCI devices and which interrupt they are assigned. Most, if not all, of the devices will have Automatic assignment of interrupts, indicating that the system will automatically assign an interrupt to that device. The P390 Adapter is probably set for 'Automatic'. Using the right and left cursor keys, change this to the interrupt you have identified or made available. Once it is changed, exit from the setup utility, making sure you Save Settings before exiting from the main menu screen.

It is possible that the automatic assignment of interrupts could still cause a shared condition. To check, repeat the steps in "Checking For Shared Interrupts" to make sure the P390 Adapter has it's own, unshared, interrupt. In some cases, it may be necessary to set all PCI adapter interrupts 'manually' (not allow the 'automatic' assignment to be used).

SCSI Configuration and Cabling

The following information is actually of primary use in confguring and ordering a PC Server S/390. However, due to the many SCSI related problems Business Partners and customers have encountered, we are providing this information here also. If you, or your customer, are having any kind of SCSI problems, review this information carefully.

SCSI Configuration

The PC Server 330 provides a very powerful and flexible base for a broard range of high speed I/O devices. This very flexibility, however, causes some complications which must be considered when configuring the PC Server 330 I/O subsystem, especially for use as a System/390. This section will address some of these considerations and suggest some alternatives to best meet most common objectives.

PB0 SCSI Configuration Alternatives

Note that the standard, default, configuration has the CDROM addressed as SCSI device 6, the internal Hot/Swap drives addressed as 0 - 5, and the controller as 7. This does not leave any address available for the 4mm DAT tape unit.

Replug H/S backplane to High-SCSI (Preferred)

• Jumper is a bit difficult / tedious to get to.
• Option and process is not documented in PC Server 330 manuals
• Is not recommended by PC Server technical support

Limit Internal Diskdrives to 5

• Limits internal capacity to 5 drives
• Performance: CD and DAT are same SCSI channel as disks

Add SCSI adapter:

• Cost of SCSI adapter
• Uses up a potentially valuable and scarce PCI slot

Use Integrated Controller for CDROM & DAT

• SCSI adapter is no longer available for external devices
• Is not recommended or supported by PC Server Technical support

Do not include / use 4mm DAT

• Unable to make use of 4mm DAT flexibility and capabilities
• Cost of SCSI 3480 tape drive (eliminating the requirement for, and cost of, the DAT does offset this cost, however.)

Use external enclosure for DAT

• Expense and space requirements for external enclosure.
• Some exposures created by additional SCSI cables (connections, cable length, etc.)

Jumpering Backplane for "High-SCSI"

To set the backplane to "High-SCSI" addresses:

1. Remove the PC Server 330 cover (observe all power-off and static-electricity cautions).
2. Provide some additional light (flashlight, desklamp, etc.) to the very upper portion of the backplane.
3. Locate the jumpers at the very top of the backplane. There will be several sets of jumpers. We're looking for four sets of jumpers (two pins each) in a vertical orientation in about the middle of the backplane and the jumper block is labeled J12. The bottom two jumpers should already be jumpered.
   (If you find it a bit difficult to see and get to this jumper, the Power Supply can be temporarily removed by removing only four screws and dropping it out of position. Place the server flat on it's back, remove 4 screws at the back of the PC Server, slide the power supply up, then lift out.)
   
4. Get a suitable additional jumper. Using "long nosed pliers" or equivalent, place the jumper over the second set of pins, which is labeled HI_ID_N. The bottom three sets of pins should now be jumpered, and the top set open.

Miscellaneous SCSI Suggestions and Warn ings

Installing a ServeRAID II

• It may be desireable to use the ServeRAID II to control the inte rnal disks (to take advantage of the battery backed-up cache and 'Data Scrubbing'). If so, we recommend using a different cable (P/N 94G3988) to connect the backplane to the ServeRaid II rather than reusing the integrated controller to backplane cable. (In most installations, reusing the standard cable results in a very tight or 'streched' fit which could cause problems later.) A new/different "drive Sense Cable" (76H5399) is also usually a good idea.
• If the integrated controller is not used for the internal disks, it can still be effectively used for the CDROM and DAT.

Using 3518 External Disk Enclosures

Recommended Install/Setup Diskettes

This section has been updated and now refers specifically to the 8640-PB0 model. While most of the listed drivers and diskettes will also function properly on the ESx models, users should verify compatibility before applying new code to old systems (especially new levels of BIOS).

These files can be obtained from the following sources:

• IBM PC Company Web Pages, file download area
• IBM S/390 Division ftp site
• Adaptec Inc. Web site, service & support.

P390 Diskettes

SCSI and RAID drivers

Suggested Diskettes

P390 Diskettes:

OS/2 Warp related Diskettes:

CHKDSK Upgrade Diskette:

Warp Server Installation Diskettes:

General PC Server 330 Diskettes

PC Server 330 (PB0) BIOS Flash Update, V 18 (or greater ):

PC Server 330 OS/2 Video Disk

PC Server 330 Ethernet MPTS Drivers, V 3.1.1:

PC Server 330 PB0 SCSI support Diskette, V 1.00: SCSI drivers and utilities, dated 10/09/1997, (05L1401.exe)

PC Server 330 PB0 7800 (SCSI) Manager:

PC Server 330 (ESx) EISA configuration diskette: Note that the new PC Server (PB0) does not have any EISA slots (only PCI and ISA), so the EISA configuration diskette is no longer needed. For users with ESx systems, the latest version of the EISA configuration diskette is V 6.01. (05L1458.exe)

ServeRAID Diskettes

ServeRAID (All) BIOS Firmware Update, V 2.70:

ServeRAID (all) Driver and Utility Disk, V 2.70: "ServerRAID controller device drivers, local RAID administration and remote server services." This diskette has the latest IPSRAID.ADD and the RAID Administration programs. Dated 04/02/1998. (24L7807.exe)

ServeRAID (all) Admin and Monitoring Utility V 2.70:

ServeRAID (all) Configuration Diskette, V2.70:

ServRAID Hard Disk Update Diskette:

Obtaining Dial-In Access

1. The 8235/DIALs product is primarily a 'hardware' solution that attaches to the LAN with little or no impact on the PC Server S/390 or other systems on the LAN. There is a corresponding component which runs in the 'remote' PC.

2. The WAN3172 Device manager, running with a MultiPort model 2 adapter, will accept dial-in connections and act as an ASYNC-SDLC protocol converter. The remote workstation requires appropriate communications software, such as PCCOM.

3. The LAN3274 Device Manager, combined with the systems TCP/IP support, will allow a inbound SLIP connection to appear to the S/390 operating system as a non-SNA 3270 device.

8235 'Dial In Access to Lans' (DIALs)

WAN3172 Device Manager

Connecting Via SLIP

This document cannot provide full details, but in general:

1. Configure the PC Server S/390 with 3270 sessions using LAN3274: Use the F2 (Update devices) and F12 (specify parameters) keys in the P390 Configurator to specify the /PORT=7490 parameter for the LAN3274 device manager.

2. IPL the 390 system and vary on/enable the 3270 address(es).

3. Enable the SLIP line by issueing the following 2 commands from an OS/2 window.

   • MODE COM3:19200

   • SLIP -COM3 -connect "ATANS.CMD"

4. Configure a SLIP connection on the client TCP/IP workstation

5. Dial/Connect from the client to the PC Server S/390, establishing the SLIP connection.

6. Open a TN3270 window and connect to the host, specifying port 7490.

7. The PC Server S/390 S/390 Logo should appear.

AWSICP (AWSPBS) VTAM Tuning Tips

This contribution was provided by John Burr of P/390 Development. It primarily addresses VTAM performance and tuning issues on a system emulating an ICA by using a "Wide Area Connector" and AWSICA device manager. It may be of some use and value for customers using a MultiPort adapter and AWSPBS.

SDLC ICAs driven by VTAM use never-ending channel programs to poll the PU2 device (e.g. 3174). As the polls and responses are short (4 bytes, including CRC) SDLC RR frames, SDLC ICAs tend to present the "heaviest" load on the system when the link is technically "idle" (and just polling). Unfortunately, this is most of the time so running a large number of SDLC ICAs can consume a large portion of the OS/2 Pentium cycles available to the P/390 channel.

There are two main VTAM tuning knobs which affect this. Both are specified on the VTAM LINE macro.

PAUSE

The ICA will "pause" during the never-ending channel program. So, if you increase this value, the channel program will not run as fast, and the P/390 channel will not be so burdened by the polling. However, if you increase it too much, users' response time will be negatively affected. I would suggest first trying PAUSE=0.1 (to see if that helps, due to the old aforementioned VTAM bug). If that doesn't help, then try PAUSE=0.2. Try this on all of the ICA lines at one time.

SERVLIM

This parm controls how often the PAUSE above occurs. The default says to PAUSE after every 4 idle polls. To cut channel load, *decrease* this value. A SERVLIM of 1 says to PAUSE after every poll and would provide the maximum efficiency. As for PAUSE, this value may negatively impact performance. I would suggest trying it this way:

1. PAUSE=0.1, SERVLIM=4
   Should be the default...do this to confirm/deny whether the PAUSE=0 bug has been hurting you and to provide a a baseline for the other tests.

2. PAUSE=0.2, SERVLIM=4
   Are things any better?

3. PAUSE=0.1, SERVLIM=1
   Probably the most you can try, without totally slowing down the users' response time. Even this *may* be noticeable. PAUSE=0.2, SERVLIM=1 is probably going too far although, if desperate, is worth a try.

:               To:                 Change in     Change in
PAUSE/SERVLIM/CPU   PAUSE/SERVLIM/CPU   Pentium CPU   Link response
-------------------------------------------------------------------
0.1  / 4     /17%   0.0  / 4     /28%   +11%          -0.1 second
0.1  / 4     /17%   0.2  / 4     /11%   - 6%          +0.2 second
0.1  / 4     /17%   0.1  / 1     / 9%   - 8%          +0.3 second
0.1  / 4     /17%   0.2  / 1     / 7%   -10%          +1.4 second

My conclusion from this study was that the "sweet spot" for efficient Pentium consumption with acceptable response time was a PAUSE/SERVMLIM of 0.1/1. So, if there's no time to diddle, go directly to those settings on all lines in the configuration and see if that buys enough to make the system run acceptably.

Optimizing Performance

There are many aspects to configuring and tuning PC Server S/390 systems to provide optimun performance. This item will attempt to address some of these considerations.

PC Memory Considerations

OS/2 Cache

Using the HPFS386 file system, you can define large OS/2 cache sizes. (Without HPFS386 the largest OS/2 cache size available is two megabytes.) The HPFS386 cache size is specified in the HPFS386.INI file in the IBM386FS directory. The parameters in this file are explained within the file, but the primary ones of interest for this discussion are CACHSIZE and USEALLMEM.

To provide best performance, the large amount of PC memory should be allocated to the HPFS386 Cache. To determine the maximum memory available, use the following guidelines:

• Assume OS/2 requires a minimum of 40MB to run effectively. So, on a base 64MB system (PB0) allocate a maximum 24MB to the cache.
• Out of each additional 32MB memory increment, allocate 20MB for the cache, and leave 12MB for OS/2.
• If any other non-S/390 related workloads are to be run on the OS/2 system, provide enough PC memory to allow that workload to run without causing swapping.

Recommendations:

Note: Users considering HPFS386 should keep im mind that the use of HPFS386 is offically not supported by P/390 Support. If you do choose to use HPFS386, we recommend the following actions.

• Apply Warp Ver 3 Fixpack 35 and Warp Server Advanced Fixpack IP_ 8508
• Set and monitor the CACHESIZE and MAXHEAP values in HPFS386.INI For CACHESIZE, use the guidelines above. For MAXHEAP, use 4000.
• If you experience any performance problems, recheck these values first. They can be changed by the automatic 'Tuning Assistant'.

S/390 Expanded Storage

Depending on the workload and S/390 operating system, Expanded Storage sizes of 64-256MB may be desireable. Since PC Server memory can be easily added and configured as Expanded Storage, it may be practical to start at a small (64MB) size and expand as workload and performance indicates. Memory reserved for ES is of course in addition to the memory used by OS/2 and any OS/2 (HPFS) caches.)

PC Server S/390 Array considerations

Array stripe unit size

On in the disk arrays of the PC Server S/390, the customer sets the stripe unit size (amount of data written on a given disk before writing on the next disk). The default stripe unit size is 8K. Choices are 8K, 16K, 32K, and 64K. Sizes larger than 8K will probably yield better performance for S/390 workloads than the default 8K.

Also consider the I/O characteristics of any other OS/2 applications that you may run concurrently on the PC Server S/390 when choosing a stripe unit size. For example, larger stripe sizes may not be the best performing choice for LAN file serving workloads. A compromise between larger and smaller stripe sizes might be appropriate depending on the overall system I/O characteristics.

Warning: Once the stripe unit is chosen and data is stored in the logical drives, the stripe unit cannot be changed without destroying data in the logical drives.

Write Policy

There are two choices for write policy with the ServeRAID adapter. The default write policy is write-through (WT), where the completion status is sent after the data is written to the hard disk drive.

To improve performance, you can change this write policy to write-back (WB), where the completion status is sent after the data is copied to the RAID adapter's cache memory, but before the data is actually written to the storage device. There is 4MB of cache memory of which more than 3MB are available for caching data.

Warning: If you use lose power before the data is actually written to the storage device, data in cache memory is lost. See also section "LAZY writes" for related information.

You can achieve a performance improvement by using WB, but you run a far greater risk of data loss in the event of a power loss than if you use WT. An uninterruptible power supply (UPS) can help minimize this risk and is highly recommended for this reason and for the other power protection benefits it supplies as well.

Users who are very concerned about data integrity (with or without UPS) and who want to take advantage of WriteBack caching may want to consider using the ServeRAID II adapter. It has an option for an 8MB "hp2.battery backed-up cache. Using the ServeRAID II for all (internal and external) disks can improve performance while providing very high levels of data integrity.

OS/2 CONFIG.SYS tuning

MAXWAIT

FAT DISKCACHE

If your PC Server S/390 has no FAT formatted partitions, then the DISKCACHE= device driver can be commented out (REM) of the PC Server S/390's CONFIG.SYS in order to save some memory. By default, OS/2 places this device driver in CONFIG.SYS. The size of the DISKCACHE may be tuned. Enter HELP DISKCACHE for information on the parameters that may be specified on DISKCACHE.

Priority_disk_IO

Specifying PRIORITY_DISK_IO=NO is recommended. NO specifies that all applications (foreground and background) are to be treated equally with regard to disk access. The default is YES. YES specifies that applications running in the foreground are to receive priority for disk access over applications running in the background.

OS/2 HPFS CACHE

Base OS/2 system HPFS Cache size

The HPFS.IFS device driver delivered with the base Warp Server (not Advanced) has a maximum cache size of 2048K (2 Megabytes). The /CACHE:nnnn parameter of the IFS device driver specifies the size of the cache. The default is 10% of available RAM (if not specified) with a maximum of 2048K. The specified value after an install of OS/2 is dependent on installed RAM at the time of installation. If you are using the standard OS/2 provided IFS device driver, then specifying /CACHE:2048 is highly recommended. Enter HELP HPFS.IFS at the OS/2 command prompt for further explanation of the parameters.

/CRECL on IFS HPFS cache

The /CRECL parameter of the HPFS IFS driver allows you to specify the size of the largest record eligible for this cache. The OS/2 default is 4K. From a S/390 perspective, increasing this value may increase cache read hits if the S/390 operating system is performing repetitive I/Os of the same data in blocks bigger than the default 4K. You can use performance analysis tools for each S/390 operating system to understand the characteristics of I/Os that are being performed by the S/390 operating system and applications. Most S/390 workloads will benefit from a CRECL value of 64 (64K). If the PC Server S/390 is using CKD disk emulation, CRECL should definitely be set to 64(k) since the AWSCKD device manager reads and writes a full track (49K) of data at a time.

Enter HELP HPFS.IFS at the OS/2 command prompt for further explanation of the parameters.

LAZY writes

Lazy writes are defaulted to ON with OS/2's HPFS. If lazy writes are enabled then when a write occurs for a block of data that is eligible for the HPFS cache, the application is given completion status before the data is actually written to the hard drive. The data is actually written to the hard drive during idle time or when the maximum age for the data is reached. Lazy writes are a significant performance enhancement.

Warning: There is a risk to the data in the event of a OS/2 software failure or power loss before the data is written from the cache to the hard drive. See section "Write Policy " for related information. You can control whether lazy writes are enabled or not with the OS/2 CACHE command (or the CACHE386 command if using HPFS386) as well as maximum age and idle times for the disk and cache buffers. Enter HELP CACHE at the OS/2 command prompt for further information. (Enter CACHE386 ? for help with CACHE386.)

DASD device drivers

Functional differences

The AWSCKD device driver has some functional differences when compared with the AWSFBA device driver. The AWSCKD device driver reads and writes a full track when an I/O is performed. The device driver has an internal cache where the track is kept until it must be flushed. As the AWSFBA device driver does not implement an internal cache, the performance characteristics between the two can be different depending upon the I/O workload. VM/ESA ESA Feature's block paging methodology seemed to benefit from the internal cache of the AWSCKD device driver in controlled laboratory experiments. You should consider using 3380 volumes for VM/ESA ESA Feature paging volumes for this reason.

You should not generalize this observation into a statement that AWSCKD performs better than AWSFBA. In fact, AWSFBA dasd volumes performed extremely well in laboratory experiments and offer some benefits over AWSCKD including finer granularity on OS/2 file allocation sizes and a close mapping to the underlying sectors of the dasd media. VM/ESA and VSE/ESA utilize FBA dasd in a very efficient manner. The flexibility of the PC Server S/390 in supporting both CKD and FBA emulated volumes in a mixture allows you to easily have both types in your configuration.

TCP/IP Configuration Information

Judging from the number of outside questions we are receiving, some elaboration about TCP/IP, LCS3172, and Ethernet seems appropriate. While the primary thrust here is Ethernet, some Token Ring considerations are also provided.

For the purposes of this discussion, one doesn't need to differentiate between VM and MVS since both are configured in a similar manner.

LCS3172 (the 3172 TCP/IP manager) links HOST (VM/ESA, VSE/ESA, MVS/ESA, or OS/390) TCP/IP software to OS/2 LAN software (LAPS (LAN Adapter and Protocol Support) or the LAPS component of MPTS (Multi-Protocol Transport Services)). There are two interfaces (HOST and LAPS) involved. Each interface must be configured to be consistent with the other. The MPTS configuration must also be consistent, at the Ethernet IP frame level, with the MPTS configuration of workstations you need to communicate with.

Fortunately, there are only a couple of parameters to configure, at least with regard to aspects of the configuration governing the behavior of LCS3172 and MPTS.

The MPTS configuration ...

• One of the differences between LCS3172 and a 3172 is that LCS3172 uses LAPS (or the LAPS component of MTPS) to interface with the Ethernet adapter, whereas the original 8232/3172 interfaced directly to an Ethernet device driver.

• Modern versions of LAPS use NDIS (Network Device Interface Specification) to interface to an NDIS Ethernet driver (usually supplied by the Ethernet adapter manufacturer). The MPTS/LAPS application interface to which LCS3172 is written is independent of the underlying network, whether it be Ethernet, Token Ring, FDDI, etc. The LAN independent nature of the LAPS interface has the following four implications for users of the PC Server S/390:

  1. Any network specific knowledge must be imparted directly to LAPS by LAPS (since, by definition, it cannot come from LCS3172).

     The MPTS/LAPS parameter which you need to be most concerned about is what frame format (DIX or IEEE 802.3) to use on your Ethernet. The frame format is set when selecting the "type of Ethernet driver support" parameter in the "IBM IEEE 802.2" protocol section of the LAPS adapter configuration.

     Much of the IEEE 802.X standard was underwitten by IBM, so most IBM software, including LAPS, defaults to the IEEE 802.3 ("I") setting. Most users, however, use the alternative standard, "DIX". You should always start your problem resolution by changing the parameter to "D" (for DIX).

     Contributing to the confusion is the fact that Novell LAN administrators are used to seeing 802.3 in Novell configuration files. But, this is used in the context of tolerating the IEEE 802.3 frames as opposed to operating in the IEEE 802.3 frame format. This has been at the root of more than one customer's insistence that they absolutely, positively were running an IEEE 802.3 Ethernet network when they weren't.

     Instances where you can observe traffic is flowing (via lights or via the LCS3172 watch window) yet nothing answers pings (and sometimes snifters cannot see any frames) are almost always due to this parameter being set incorrectly to "I" (IEEE 802.3).

     Although other things like duplicate IP addresses can cause similar problems, you should change this parameter to "D" before changing or adjusting any other parameter. Only if you are very, very sure you know what you are doing, should you change "type of Ethernet driver support" to "I" (IEEE 802.3) (or accept the default).

     Token Ring does not have a similiar problem because there is only one frame format standard (IEEE 802.5).

  2. Certain aspects of Ethernet frames and frame sizes must be considered. The LAPS application interface is designed to handle frames for any type of network. In practical terms, this means that the interface is essentially in Token Ring (IEEE 802.5) format. In other words, an interface able to handle a Token Ring frame can handle an Ethernet frame (but not the other way around).

     The original 8232/3172 Ethernet used a device driver written for the only Ethernet adapter then available. Also, the HOST TCP/IP code, when configured for Ethernet, formatted frames in exactly the format required by that Ethernet adapter device driver.

     With LCS3172, Ethernet frames are accepted from the HOST in Ethernet format, are mapped to LAPS (Token Ring) format, and passed to LAPS for transmission. This mapping adds a SNAP DLC header and usually reduces the maximum length of a frame that could otherwise have been transmitted. Also, different NDIS driver implementations may calculate the maximum Ethernet frame size differently. The worst case combination of these two reductions has not yet exceeded eight bytes.

     So, if you would ordinarily specify 1500 bytes in the HOST TCP/IP configuration file, reduce it to 1492 bytes and run LCS3172 with the watch text window (change "AWSSTART LCS3172.EXE N" to "AWSSTART LCS3172.EXE T" in IPL.CMD). As long as there are no "(XSIDE ERROR) DIR.TRANSMIT.FRAME INVALIDLENGTH" errors reported, then the value for maximum frame size is acceptable.

     Again, Token Ring users will not be subject to a difference in maximum frame size, since the frames sent from the HOST TCP/IP are in exactly the format required by LAPS for transmission.

     Users of the LAN Streamer Token Ring adapters may notice HOST TCP/IP messages concerning rejection of frames with a route length field exceeding 18 bytes. The routing field length might be a problem only on networks on which a frame could pass through more than eight bridges. This is because previous Token Ring (shared memory) adapters were limited to an 18 byte routing field. But, Streamers are able to handle routing fields to the architected length (which is more than 18 bytes). Since later 3172 models allow LAN Streamer Token Ring adapters and would presumably receive these frames as well, a fix for HOST TCP/IP may be required.

  3. To run TCP/IP for OS/2 concurrently with LCS3172, then you will need two different LAN adapters (or one of the dual LAN adapters). The usual LCS3172 conflict is with TCP/IP for OS/2 which has the same IP layer requirement (and the same need for the LAPS DIRECT interface). Since both LCS3172 and TCP/IP work at the Ethernet IP layer, both of them must use the LAPS DIRECT interface. The LAPS DIRECT interface can be considered a special (SAP=0) case of the general LAPS 802.2 DLC interface. A Service Access Point (SAP) is the mechanism by which different OS/2 applications share the same LAN adapter. Since the SAP is the sharing mechanism, it follows that only one OS/2 application can have the same SAP open at a time. Specifically, only one OS/2 application can use the LAPS DIRECT (SAP=0) interface at a time.

     IBM Token Ring users may also note a conflict with TRMON (an IBM internal Token Ring monitor utility) which also uses the LAPS Direct interface.

  4. LCS3172 can be used with any LAN adapter for which there is an OS/2 LAPS NDIS compliant driver available.

The HOST (VM and/or MVS) configuration

• Ethernet (DIX) frame formats are passed between HOST and LCS3172

• LCS3172 and LAPS exchange the frames in the LAPS network independent format

• LAPS presents/interprets Ethernet frames according to how LAPS is configured.

The Bottom Line

1. Change the LAPS default value for "type of Ethernet driver support" from "I" to "D"

2. Code ETHERNET in the HOST TCP/IP LINK statement and use 1492 as the maximum frame value on all of the HOST TCP/IP statements that require it in reference to that Ethernet link. Naturally, if your Ethernet topology limits you to a value smaller than 1492, then code that smaller value.

3. If there are still problems, you should then go ahead and try the easy fix and change the LAPS "type of Ethernet driver support" parameter to "I". Although this is unlikely to correct your problems, it is an easy change to make and test.

4. If there are still problems, insure the LAPS "type of Ethernet driver support" parameter is set to "D" and turn your attention away from LCS3172 and LAPS. Then focus on your Ethernet topology (ie., away from the HARDWARE definition section of the HOST TCP/IP configuration file and towards the HOME and GATEWAY sections of the HOST TCP/IP configuration file).