Previous Topic: Evaluate Performance Using the Investigate DisplaysNext Topic: Investigate an I/O Limitation


Investigate a Memory Limitation

The chief indicators of a memory limitation are as follows:

No free memory

Look at the value of free pages, the memory bar histogram, and the memory queue (Mem que).

If the value of free pages approaches that of FREELIM (the system parameter that sets a lower limit on the number of pages on the free list), this indicates a memory shortage.

The memory bar histogram shows the amount of memory used by the free and modified page lists with respect to that available for user working sets. Again, a shortage of memory for the page caches (free plus modify lists) indicates a high degree of user memory utilization, and consequently a memory shortage.

A nonzero value of Mem que indicates processes in the memory queue in computable outswapped states, awaiting memory that is unavailable.

Page fault high

Look at the value of Pgflts (page faults), Sysflts (system faults), and paging in the IO_window for disks.

For a VAX-11/780 CPU, a value of Pgflts (page fault rate, or number of faults [hard and soft, including system] per second) greater than 100 is cause for concern. For other CPUs, use an appropriate threshold such as that supplied with the factory knowledge base. If Pgflts is greater than this number, page faults might be excessive on your system.

Sysflts (the rate at which pages are faulted into the system working set) should be no more than 1 fault per second; otherwise, the system is faulting itself to do work on the users' behalf. If system faults are high, it might be necessary to increase the value of system parameter SYSMWCNT, which controls the system working set size.

If disks are spending an excessive percentage of their time doing paging (indicated by the disk bars in the IO_window), then a memory limitation is causing harmful I/O effects.

Swapping high

Look at the value of “Inswaps,” the value of “Mem que,” and the amount of swapping and modified page writing done by the disks (Swap/Mod in the disk bars).

Inswaps gives the number of processes that were swapped back into memory during the last sample interval, and Mem que is the number of processes waiting for memory. If these are significant, then swapping is a problem on your system, and either indicates a memory limitation or memory management problem.

If disks are spending an excessive percentage of their time doing swapping and modified page writing (indicated by the disk bars in the IO_window), then a memory limitation is causing harmful I/O effects.