|Home | Products & services | Support & downloads | My account|
PC Server 500 System/390 Performance
Copyright © IBM Corporation, 1995. All rights reserved.
This document was originally published in June 1995 as WSC Flash 9522 docid G023542 in IBM's FLASHES database. It has been converted to HTML for display here. No change in content has been made except for the addition of alternate table views for those web browsers that cannot process tables correctly and links to online versions of documentation cited in the original flash. Documentation for VM/ESA performance on the follow-on product to the PC Server 500 S/390 can be found in a document named VM/ESA® Peformance on P/390 and R/390 - PC Server 520 and RS/6000 591 at http://www.vm.ibm.com/perf/docs/p520r591.html. Additional related information can be found at http://www.s390.ibm.com/products.
WSC Flash 9522. Document id G023542
The information contained in this document has not been submitted to any formal IBM test and is distributed on an as is basis without any warranty either expressed or implied. The use of this information or the implementation of any of these techniques is a customer responsibility and depends on the customer's ability to evaluate and integrate them into the customer's operational environment. While each item may have been reviewed by IBM for accuracy in a specific situation, there is no guarantee that the same or similar results will be obtained elsewhere. Customers attempting to adapt these techniques to their own environments do so at their own risk.
Performance data contained in this document were determined in various controlled laboratory environments and are for reference purposes only. The results that may be obtained in other operating environments may vary significantly. Users of this document should verify the applicable data for their specific environment.
References in this document to IBM products, programs or services do not imply that IBM intends to make these available in all countries in which IBM operates. Any references to an IBM licensed program in this publication is not intended to state or imply that only IBM's programs may be used. Any functionally equivalent program may be used instead.
This flash contains tuning and performance details for the PC Server 500 System/390 system that augment the information provided in the document entitled IBM PC Server 500 S/390 ...Is it right for you? Technical Application Brief available from MKTTOOLS in PCSVR390 PACKAGE or orderable as publication number GK20-2763. The publication is not scheduled to be orderable using the publication number until general availability of the IBM PC Server 500 S/390 in July 1995.
Information in this flash was developed jointly by the performance department at the Endicott Programming Lab, PC Server S/390 Competency Center in Atlanta, and S/390 Performance Design in Poughkeepsie. Questions may be directed to RHQVM15(PCSVR390) or PCSVR390@VNET.IBM.COM.
Other information sources are
The IBM PC Server S/390 system is a combination of the System/390 and PC Server architectures that provides access and use of both in a single package. While the S/390 instructions execute natively on a dedicated CMOS chip on the S/390 Microprocessor Complex, the execution of the S/390's I/O is handled by OS/2 device managers, device drivers, and S/390 channel emulation. The S/390 design point in the PC Server S/390 is unique when compared to other S/390 processors. In this implementation, S/390 devices (tapes and printers) are either channel attached (via S/370 Channel Emulator/A) or emulated on PC devices in a manner that is transparent to the S/390. In addition to emulating the S/390 I/O, the high performance Pentium® processor also supports OS/2 applications and advanced local area network (LAN) functions.
While this mix creates an environment with many exciting options for I/O attachment and support, some S/390 capabilities are not available with this technique. For example, the I/O design point for the PC Server S/390 does not support multiple channels attached to the same physical device, a capability that is exploited in other S/390 systems.
Application of the PC Server S/390 to a specific customer's unique workload requires some planning to achieve good results. The contents of IBM PC Server 500 S/390 ...Is it right for you? provide the essentials for determining if there is a possible match between this system and the targeted work. This flash provides more detail that can be used to fine tune applications of this system to customer environments.
The following is a list of highlighted points that are summarized from the text of the flash:
Two different server hardware configurations were used in the
laboratory performance measurements for VM/ESA and VSE/ESA. The hardware
descriptions are in Table 1. If
your browser cannot display tables correctly, then click here for an
alternate view of Table 1.
TPNS measurements were obtained by utilizing the AWS3172 device driver and establishing a VTAM 3172 XCA connection across an isolated Token Ring LAN between a PC Server 500 S/390 running VM/ESA ESA Feature with VTAM and a second server running VM/ESA 370 Feature with VTAM and IBM's Teleprocessing Network Simulator product (TPNS). TPNS simulates actual VTAM cross domain logon sessions to the target system. The simulated VTAM sessions logged onto the VSCS APPL on the target system. CMS users were then logged onto on the measured system, and the users ran their scripts with TPNS measuring the end user response time through the VTAM network. Thus, the CMS workload measurements also stressed the LAN adapter on the server as the VTAM data streams flowed back and forth across the Token Ring. The setup is pictorially represented as follows:
*-----------------* | PC Server S/390 | *-------* | VM/ESA ESA *---------* 8228 | Isolated | VTAM/VSCS | | M.A.U.| Token Ring | | *---*---* *-----------------* | Measured System | | | | *-------*-------------* | PC Server | | VM/ESA 370 Feature | | VTAM | | TPNS | *---------------------* TPNS Driver System
This section contains a summary of the performance measurement results that have been obtained for VM/ESA, VM/ESA + LAN File Serving, VSE/ESA, and MVS/ESA.
Several measurements were obtained using the FS8F workload to get an understanding of how many CMS users can be supported in various S/390 storage sizes, while maintaining an average response time of (approximately) one second. The results are summarized in Table 2.
FS8F represents a CMS program development environment. It includes a wide range of CMS and Xedit commands along with assembler, COBOL, FORTRAN, and PL/I compiles and program executions. Users are simulated as TPNS scripts. The average think time of 26 seconds and the think time distribution are intended to reflect the average characteristics of a logged-on CMS user. The minidisk-only version of FS8F was used for these measurements. See Appendix A of VM/ESA Release 2.2 Performance Report, GC24-5673-01, for further information.
The following conditions applied to all three measurements:
In the following table and elsewhere in this section, System 1 and
System 2 refer to the two measurement configurations described in Table 1. If
your browser cannot display tables correctly, then click here for an
alternate view of Table 2.
The results show that, for this workload, the number of CMS users that can be supported is mostly determined by the amount of S/390 memory that is available. Contention for the S/390 processor is not a significant factor until 128MB are made available, at which point S/390 processor utilization rises to 80%.
VM/ESA performs better on S/390 when steps are taken to minimize the amount of page I/O that the system has to do. Page I/Os are expensive because each I/O typically reads or writes multiple 4K pages. In addition, VM/ESA's block paging mechanism is optimized for traditional mainframe DASD and therefore does not work as well with the device emulation used here. We reduced the page I/O rate by taking the following tuning actions:
We also found that the response time impact of VM/ESA page I/O tends to be reduced when emulated CKD devices are used as paging volumes and multiple page volumes are defined. This does, however, tend to result in higher Pentium® utilizations relative to using FBA page volumes.
These measurements were done with write caching in effect. An additional measurement (not shown) was done without write caching. The results showed that the number of CMS users had to be reduced by 30% in order to maintain 1-second average response time. See section "VSE/ESA: CICS transactions" for additional discussion of write caching.
The Pentium utilization arises from handling the S/390 I/O requests. As this utilization increases, contention for the Pentium will cause I/O service times (as seen by the S/390) to increase. The Pentium, then, is one of the resources that can limit the S/390 I/O rate that can be sustained while still providing acceptable response times.
Note that the OS/2 I/O rate (as reported by SPM2) is higher than the S/390 I/O rate (as reported by VMPRF). This is because the S/390 I/O emulation code will sometimes split one S/390 I/O request into multiple OS/2 I/O requests.
The 64MB results should be viewed as approximate. This is because 1) the measurement was done on a system with 128MB with VM/ESA generated to only use 64MB and 2) the measurement was done using the System 1 configuration. We expect that a measurement done with the System 2 configuration and 64MB would be similar, but with a lower average response time (about 1 second) and a lower Pentium utilization.
The three measurements summarized in Table 3 illustrate the interactions that can occur when VM/CMS work on the S/390 is combined with LAN file serving work. The first two measurements show the effects of doing each type of work alone in a dedicated manner and the third shows the effects of running the two workloads simultaneously.
The following configuration and set-up parameters apply to all three measurements:
Except for the size of the HPFS386 cache (which was changed to 17M), OS/2 LAN Server was not tuned. The VM system was tuned in the same manner as all other runs. These tuning parameters may not represent optimums for the combined environment.
No significance should be assigned to the relative amount of work that was being run in each of the dedicated runs. There will be different relative amounts of work for each customer environment with potential for combined work. However, since each workload contends for many of the same resources, it is expected that all environments with combined work will show some degree of interaction. The amount of interaction will be dependent on many factors such as I/O rates, cache sizes, storage size, and the specific type of and of amount work being run in each environment.
The table shows that the combination of the two workloads caused the response time of the CMS users to increase by about 0.4 sec while the response time to the LAN clients increased by about 7% over their respective dedicated values. The Pentium utilization per I/O is much less for the LAN environment as compared to the S/390 I/O support and is mainly due to the processor cycles required for the S/390 I/O emulation. The CMS environment was impacted more than the LAN work by the relatively higher demand that the LAN serving work placed on the disk adapter/array. In this respect the arbitrary amount of LAN work chosen for the measurement was more intense than the CMS work.
If your browser cannot display tables correctly, then click here for an
alternate view of Table 3.
A number of measurements were obtained for VSE/ESA native and VM/ESA guest environments using the VSECICS workload to get an idea of how much CICS on VSE transaction processing can be done on the PC Server 500 S/390 under various conditions. A subset of these results is summarized in Table 4.
The VSECICS workload consists of seven CICS applications, written in COBOL and assembler, which include order entry, receiving and stock control, inventory tracking, production specification, banking, and hotel reservations. These applications invoke a total of 17 transactions averaging approximately 6 VSAM calls (resulting in about 3 DASD I/Os) and 2 communication calls per transaction. Terminals are simulated by an internal driver tool. See Appendix A of VM/ESA Release 2.2 Performance Report, GC24-5673-01, for further information.
Except where noted in the results tables, the following conditions apply to all the VSECICS measurements:
For the guest measurements:
These measurements were done with 32MB of S/390 memory. This was more than adequate for this workload and there was negligible paging.
These results show guest-to-native CPU usage ratios that are similar to those observed on mainframe processors for this workload. The guest-to-native ratio is influenced by the I/O content of the workload. Workloads that are more I/O-intensive will tend to show higher (less favorable) CPU usage ratios, while workloads that are less I/O-intensive will tend to show lower CPU usage ratios.
Regarding the V=R case, the PC Server 500 S/390 does not provide I/O passthru (SIE assist). It does, however, support the CCWTRANS OFF optimization.
These measurements were obtained using the System 1 configuration. The System 2 configuration would have yielded somewhat lower response times (faster DASD, larger array stripe width) and lower Pentium utilizations (the 3/95 390 LIC level does more efficient S/390 I/O emulation).
The default write cache settings were used for these measurements. That is, write caching was specified for OS/2's HPFS cache (lazy on) while write caching was not specified for the RAID adapter cache (write-through). An additional measurement (not shown) was obtained with write caching done in both the HPFS cache and the RAID adapter cache (lazy on, write-back). This showed little or no additional improvement relative to lazy on, write-through. Another measurement was taken with write caching done only by the RAID adapter cache (lazy off, write-back). This yielded very similar results to the lazy on, write-back results shown above. Taken together, these findings indicate that 1) write caching is equally effective at improving performance, whether done in the HPFS cache or the RAID adapter cache and 2) it is only necessary to do write caching in one of these two caches to achieve all or most of the write caching benefits.
As discussed in the tuning section, the safest way to do write caching is to use the RAID adapter's cache configured for write-back with the system protected by an uninterruptible power supply with software caching by OS/2 turned off (LAZY=OFF).
When write caching was used, response time improved by 0.4 seconds even though the number of users was increased from 140 to 225.
The VSECICS workload has an exceptionally low read/write ratio (0.6). Workloads having higher read/write ratios would receive correspondingly less benefit from the use of write caching.
For this workload, when write caching is used, system capacity is gated by S/390 processor speed. When write caching is not used, system capacity (given the 1-second response time goal) is gated by the I/O subsystem.
The write caching measurement was done using HPFS write caching (lazy on). Similar results have been observed with RAID adapter write caching (write-back).
The total disk I/O throughput on this platform can be significantly improved by paying particular attention to the array and channel configuration of the PC Server S/390. Just as in the mainframe world, spreading I/O operations over multiple drives and S/390 channels improves potential throughput. This is accomplished on the PC Server S/390 by creating multiple arrays and spreading the data across multiple RAID adapter channels. The IBM SCSI-2 F/W Streaming-RAID Adapter/A has two SCSI channel connectors on the adapter and 4MB of cache of which more than 3MB are available for caching. The cache may be configured either Write Through or Write Back.
Laboratory measurements were performed in three different I/O configurations using a VSE batch workload.
Each array contained three 2.25G drives configured into an array with a
RAID-5 logical drive. All S/390 DASD volumes in all three configurations
were emulated FBA devices. If your browser cannot display tables
correctly, then click here for an
alternate view of Table 6.
In I/O Config 2, the data volumes were spread across two arrays and two channels, the I/O overlap increased and allowed an increase in I/Os per second executed by 31%. Taking this a step further, the same workload was run with two arrays and two RAID adapters (one channel each) and achieved 61% improvement in I/O rates over the base configuration (one array and one channel). Adding the second adapter provided not only the additional channel, but also an additional 3MB in write back cache. The OS/2 files containing the S/390 DASD volumes that were placed on the second array were chosen so that the I/O in the workload was balanced across the arrays. This action is comparable to tuning by spreading the I/Os across mainframe DASD volumes and channels.
Pentium capacity limits the extent to which such extensions to the I/O subsystem can increase I/O capacity. For the I/O Config 3 case, Pentium utilization has risen to 81%, indicating that Pentium capacity is coming close to being the limiting resource.
Caution: The I/O rates shown in this section were obtained using a batch workload. The I/O rates that can be achieved by interactive workloads, while maintaining adequate response times, will typically be much lower.
It should be noted that occupying Bank D of the PC Server 500 requires installation of a DASD backplane (into which the hot-pluggable drives plug) and also the optional 220-watt power supply. The optional power supply provides power for Banks D and E.
The performance data in the table below provides guidance to help you determine if your dedicated MVS production online workloads will fit on the PC Server 500 System/390. The table includes several online environments and their key characteristics. Due to the inherent I/O content of these workloads, the disk I/O rate becomes a key factor to consider as you evaluate the potential use of the PC Server 500 System/390 in your business.
The data in the table is based on a PC Server 500 System/390 system that has 128M of S/390 storage, 2 bays of 2GB disk drives containing 11 drives total. The drives are configured into two arrays with 1 logical drive each and RAID level 5 was specified. Each logical drive contained 10 3380 (various densities) equivalents and the 3380s were loaded with the workload components so that the I/O rates to each logical drive were no more than 60 percent of the total I/O rate. A smaller system with a single array on a single channel will handle about 1/2 the I/Os and users shown. All workloads have LAZY set to off for the HPFS cache and CICS also has the RAID adapter cache set to write-through. IMS and DB2 have the RAID adapter cache set to write-back, thus allowing disk writes to be cached for these two workloads.
If your browser cannot display tables correctly, then click here for an
alternate view of Table 7.
While the internal response times achieved by the PC Server 500 System/390 are within the specified limits, other S/390 processors, due to their standard I/O design point, typically yield lower response times. Since internal response time is one of several factors that contribute to end user response time. there may be some instances where end user times are longer than those achieved by other S/390 processors. There are many instances, particularly in remote applications, where use of the PC Server 500 System/390 can eliminate or reduce other time components to yield net improvement in overall response time.
Interactive users for these MVS workloads are simulated with an IBM internal driver tool and not with TPNS.
The IBM internal IMS workload consists of light to moderate transactions covering diverse business functions, including order entry, stock control, inventory tracking, production specification, hotel reservations, banking, and teller systems. These applications are similar to the CICS applications but contain IMS functions such as logging and recovery. The IMS workload contains sets of 17 unique transactions, each using a different database. The workload uses both VSAM and OSAM databases with VSAM primary and secondary indexes.
The DB2 workload consists of light to moderate transactions from two defined and well-structured applications -- inventory tracking and stock control. IMS/DC is used as the transaction manager. The applications are functionally similar, but not identical to, two of the IMS/DL1 and CICS applications. The DB2 contains seven unique transactions. Conversational and wait-for-input transactions are not included in the DB2 workload.
The CICS workload consists of light to moderate transactions from many of the same applications mentioned for the IMS work. The CICS applications are written in COBOL or Assembler and are functionally similar, but not identical, to the applications used in the IMS workload and use VSAM datasets only. There are six sets of 17 unique transactions.
For MVS/TSO performance data, please refer to IBM PC Server 500 S/390 ... Is it right for you? Technical Application Brief referenced in the Introduction section of this document.
This section contains performance and tuning hints that were gathered during laboratory measurements and experiences. It is intended to give information that may be useful in planning for installation and tuning a PC Server 500 System/390.
On array models of the PC Server 500 System/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 500 System/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 in order 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.
There are two choices for write policy with the RAID 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. 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.
The HPFS.IFS device driver delivered with OS/2 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.
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. OS/2 performance tools like IBM's SPM/2 V2 can also assist in tuning the /CRECL value.
Enter HELP HPFS.IFS at the OS/2 command prompt for further explanation of the parameters.
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, especially on non-array models of the PC Server 500 System/390 where there may be no hardware caching on the SCSI adapter.
Warning: There is a risk to the data in the event of an 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.)
Lazy writes are defaulted to ON with OS/2's FAT DISKCACHE. If lazy writes are enabled then when a write occurs for a block of data that is eligible for the FAT 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, especially on non-array models of the PC Server 500 System/390 where there may be no hardware caching on the SCSI adapter.
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 or not lazy writes occur for the FAT cache with parameters on the DISKCACHE= statement in CONFIG.SYS. Enter HELP DISKCACHEat the OS/2 command prompt for more information on DISKCACHE parameters.
If your PC Server 500 System/390 has no FAT formatted partitions, then the DISKCACHE= device driver can be commented out (REM) of the PC Server 500 System/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.
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.
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, less Pentium time to handle S/390 I/Os, 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 500 System/390 in supporting both CKD and FBA emulated volumes in a mixture allows you to easily have both types in your configuration.
Newer technology LAN adapters such as IBM's Streamer family are highly recommended for maximizing the communications throughput of the PC Server 500 System/390.
Information in this section is specific to the named adapter and does not apply to the "Streamer" family of IBM LAN adapters.
The value for XMITBUFSIZE (Transmit Buffer Size) is a tuneable value for this adapter card. The default value used for IBM's 16/4 Token Ring Adapter/A may be a poor choice if you are using VTAM for subarea communications between two VTAM subareas. When performing full screen operations such as XEDIT under VM/CMS, the buffer used by VTAM will exceed the XMITBUFSIZE size specified in PROTOCOL.INI and cause segmentation. For example, when using a 16/4 Token Ring Adapter/A in a laboratory environment, multi-second response time was observed while scrolling in XEDIT when logged on via a cross-domain VTAM session from one PC Server 500 System/390 to another. Increasing the value of XMITBUFSIZE so that it was more than the VTAM RUSIZE restored response time to its expected sub-second value. A rule of thumb for tuning XMITBUFSIZE:
z = (VTAM RUSIZE in bytes) + 9 + 40 minimum XMITBUFSIZE = Round-to-next-highest-multiple-of-eight( z )where the "9" is the nine bytes for the transmit header and the request header, and the "40" is some extra to give a little room for bytes that may not be accounted for at this time. Note that there are different maximums for XMITBUFSIZE depending on whether your token ring is a 4Mbit or 16Mbit ring. For example, the maximum size of XMITBUFSIZE for the IBM 16/4 Token Ring Adapter/A on a 4Mbit ring is 4456. Other older adapters have limits that are smaller still for 4Mbit rings.
It should also be noted that in this particular situation, when REMOTE was set ON under CMS/XEDIT, data compression performed by CMS for fullscreen I/O also restored sub-second response time. This indicates the continued value of this virtual machine setting in tuning for VTAM use in a VM environment.
The following terms, are trademarks of the IBM Corporation in the United States or other countries:
The following terms are trademarks of other companies as follows.
|About IBM | Privacy | Legal | Contact|