A number of factors affect VMS system performance, either alone or in relationship to others.

In a virtual memory operating system one of the most critical factors is the relationship between physical memory and on-disk page (swap) space. When physical memory becomes constrained (either on a per-process - in the VMS environment, or on a system-wide basis) performance becomes a function of the frequency and efficiency of paging to external storage.

Insufficient physical memory for the tasks being undertaken is indicated when a system is encountering poor or patchy responsiveness coupled with related, excessive paging to disk.

Poor performance assumes a number of forms, depending on the expectations of the user of a system, as well as the demands being placed on that system.

Sluggish Responsiveness?

As observed in an interactive context (e.g. editing), this behaviour is largely a function system load in one or combination of the following factors:

• CPU Load
• Physical Memory Load

Generally, when CPU load is high responsiveness will of course be affected to a greater or lesser degree. However if CPU usage is low or moderate sluggish response can often be attributed to insufficient physical memory (for the current job profile), as indicated by high levels of paging to disk.

Poor Throughput?

Ignoring application design issues, this behaviour is related to one or more of the following factors:

• CPU Load
• Physical Memory Load
• I/O Load

Of course, job throughput is related to CPU processing power and load. However, if job throughput is unreasonably low and CPU usage is low to moderate, these factors can affect performance:

• Excessive Paging.
  • As a result of insufficient physical memory to support the virtual memory demands of the current job profile.
  • As a result of insufficient working set quota to support the virtual memory demands of the single job. This is a per-user process quota, that can be changed by VMS systems management.

• I/O Limitations.

  I/O bandwidth (for a single job to a single mass-storage device) or I/O congestion (where multiple jobs are accessing a common mass-storage device).

More Information?

See bookreader document Guide to OpenVMS System Performance possibly available via HyperReader.

HyperSPI v1.3;  August 1998