Summary of Key Findings
This section summarizes key z/VM 5.4 performance items and contains links that take the reader to more detailed information about each one.
Further, our performance improvements article gives information about other performance enhancements in z/VM 5.4.
To compare performance of z/VM 5.4 to z/VM 5.3, IBM ran a variety of workloads on the two systems. For the base case, IBM used z/VM 5.3 plus all Control Program (CP) PTFs available as of November 1, 2007. This was the first CP that had a fully functional CMMA. For the comparison case, IBM used z/VM 5.4 plus the GA (aka first) RSU.
Regression measurements comparing these two z/VM levels showed nearly identical results for most workloads. Variation was less than 5% even for workloads that may have received some benefit from z/VM 5.4 performance improvements.
Some workloads with MDC active experience a reduction in transaction rate and increased system time caused by excessive attempts to steal MDC page frames. The reader can find more information in our MDC discussion.
Key Performance Improvements
z/VM 5.4 contains the following enhancements that offer significant performance improvements compared to previous z/VM releases:
Dynamic Memory Upgrade: z/VM 5.4 allows real storage to be increased dynamically by bringing designated amounts of standby storage online. Further, guests supporting the dynamic storage reconfiguration architecture can increase or decrease their storage sizes without taking a guest IPL. On system configurations with identical storage sizes, workload behaviors are nearly identical whether the storage was all available at IPL or was achieved by bringing storage online dynamically. When storage is added to a VM system that is paging, transitions in the paging subsystem are apparent in the CP monitor data and Performance Toolkit data and match the expected workload characteristics.
Specialty Engine Enhancements: z/VM 5.4 provides support for the new z/VM-mode logical partition available on the z10 processor. A partition of this mode can include zAAPs (IBM System z10 Application Assist Processors), zIIPs (IBM System z10 Integrated Information Processors), IFLs (Integrated Facility for Linux processors), and ICFs (Internal Coupling Facility processors), in addition to general purpose CPs (central processors). Guests can be correspondingly configured. On system configurations where the CPs and specialty engines are the same speed, performance results are similar whether virtual specialty engines are dispatched on real specialty engines or simulated on CPs. On system configurations where the specialty engines are faster than CPs, performance results are better when using the faster specialty engines and scale correctly based on the relative processor speed.
DCSS Above 2 GB: In z/VM 5.4, the utility of Discontiguous Saved Segments (DCSSs) is improved. DCSSs can now be defined in storage up to 512 GB, and so more DCSSs can be mapped into each guest. New Linux support takes advantage of this to build a large block device out of several contiguously-defined DCSSs. Compared to sharing read-only files via DASD or via Minidisk Cache (MDC), sharing such files via XIP in DCSS offers reductions in storage and CPU utilization. In the workloads measured for this report, reductions of up to 67% in storage consumption and 11% in CPU utilization were observed.
TCP/IP Layer 2 Exploitation: In z/VM 5.4, the TCP/IP stack can operate an OSA-Express adapter in Ethernet mode (data link layer, aka layer 2, of the OSI model). Data is transported and delivered in Ethernet frames, providing the ability to handle protocol-independent traffic. Measurements comparing Ethernet-mode operation to the corresponding IP-mode setup show an increase in throughput from 0% to 13%, a decrease in CPU time from 0% to 7% for a low-utilization OSA card, and a decrease from 0% to 3% in a fully utilized OSA card.
Other Functional Enhancements
z/VM 5.4 contains these additional enhancements which, though not developed specifically for performance reasons, IBM felt it appropriate to evaluate for this report.
Telnet IPv6: In z/VM 5.4, the TCP/IP stack provides a Telnet server and client capable of operating over an Internet Protocol Version 6 (IPv6) connection. This support includes new versions of the Pascal application programming interfaces (APIs) that let Telnet establish IPv6 connections. Regression measurements showed that compared to z/VM 5.3 IPv4 Telnet, z/VM 5.4 IPv4 Telnet showed -8% to +3% changes in throughput and 3% to 4% increases in CPU utilization. New-function measurements showed that compared to z/VM 5.4 IPv4 Telnet, z/VM 5.4 IPv6 Telnet showed increases from 12% to 23% in throughput and decreases in CPU utilization from 3% to 13%.
CMS-Based SSL Server: In z/VM 5.4 IBM rewrote the SSL server so that the server runs in a CMS machine instead of in a Linux machine. Regression measurements showed that compared to the Linux implementation, the CMS implementation costs more CPU to create a new connection and to send data on a connection, especially as the number of concurrent connections grows large. Measurements of new function showed that the cost to create a new connection increases with key length. Said measurements also showed that high cipher mode is more efficient than medium cipher mode, because in high cipher mode the server exploits the CP Assist for Cryptographic Function (CPACF).