Contents | Previous | Next

2064-1C8, 2G/6G

Workload: CMS1 with Internal TPNS:   (Ref #1.)

Hardware Configuration:  

Processor model:            2064-1C8
Processors used:            8
Storage:
    Real:                   2GB (default MDC)
    Expanded:               6GB (MDC fixed at 512M)
 
DASD:

Note: The DASD volumes are distributed across 13 RVA T82 subsystems. Each subsystem has 1536M of cache.

Software Configuration:  

Driver:                     TPNS
Think time distribution:    Bactrian
CMS block size:             4KB
 
Virtual Machines:

Measurement Discussion:  

This environment is sufficiently storage constrained that there is heavy paging to expanded storage and some paging to DASD.

The measurement results are summarized in the following two tables. The first table contains the absolute results, while the second table show the results as ratios relative to the VM/ESA 2.4.0 base case. RTM and TPNS log data were not collected for these measurements. When available, corresponding VMPRF data has been substituted for the RTM data normally shown.

The results show that z/VM 3.1.0 with 31-bit CP total processing requirements (PBT/CMD (H)) are equivalent to VM/ESA 2.4.0. z/VM 3.1.0 processing requirements are about 1.8% higher as the result of a 7.1% increase in CP CPU time (CP/CMD (H)).

The MASTER CP results suggest that the amount of time that CP must run on the master processor has increased somewhat. This increase is close to run variability for z/VM 3.1.0 with 31-bit CP. For z/VM 3.1.0 with 64-bit CP, the increase was 4.7%. Overall CP CPU utilization grew by 1.9% (calculated as TOTAL (H) - TOTAL EMUL (H)). The increased growth in MASTER CP beyond 1.9% represents an increased tendency to run on the master processor. This does not necessarily mean that there is more CP code that must run on the master processor. It could instead mean that work that no longer needs to run on the master processor is being moved to an alternate processor less quickly.

These results focus on migration of CP to z/VM 3.1.0. The CMS and GCS components were kept at the VM/ESA 2.4.0 level for all three measurements. This means that the z/VM 3.1.0 results do not include the effects of the Rexx runtime improvement (see Miscellaneous CMS Improvements), which reduces total CPU time per command command by about 0.2% for this workload.

Table 1. CMS-Intensive Migration from VM/ESA 2.4.0: 2064-1C8, 2G/6G

Table 2. CMS-Intensive Migration from VM/ESA 2.4.0: 2064-1C8, 2G/6G - Ratios

Contents | Previous | Next