User Guide
30 Server Memory Administration Guide
Server Memory Administration Guide
103-000147-001
August 30, 2001
Novell Confidential
Manual 99a38 July 17, 2001
Responding to Disk Thrashing
The real value of virtual memory is in using a sufficient supply of memory
more efficiently, thereby improving server performance.
If the overall supply of memory is running low, then swapping will occur more
often when virtual memory is in use. If memory is extremely low, the system
might spend a significant amount of its time swapping memory in and out of
disk and have little time to accomplish useful work. This is called disk
thrashing.
In extremely low memory conditions, NetWare moves all the memory from a
protected address space into the swap file, temporarily stopping the modules
within the space. After a period of time, NetWare moves the memory back into
the protected address space and shuts down another space, moving its memory
to the swap file. Without virtual memory, these extremely low memory
conditions would cause processes to fail. With virtual memory, the server
keeps running, although very slowly.
If the Average Clean Free Available State or the Average Page I/O state is
Suspect or Bad on the Virtual Memory Health Information page in NetWare
Remote Manager, your server is experiencing disk thrashing.
If your server is experiencing disk thrashing, you can use the following two
hidden SET parameters (Memory category) to manage the balance between
the virtural memory cache pool and the file system cache pool. Use caution
when changing these parameters, because until you have more than 4 GB of
memory in the server, increasing these parameters will remove memory from
the file system cache pool and cause low memory conditions there.
VM Cache Pool Free Desired Pages, range 0 to 2147483647, default 256
Specifies the number of pages that the viritual memory system cache
needs to operate properly.
VM Cache Pool Free Lots of Pages, range 0 to 2147483647, default 512
Specifies the number of pages that the virtual memory system cache
considers to be excessive. Having an available page count higher than this
number might result in the pages being be given to another cache pool.
You can specify the point at which the server sends an alert to the console
because of excessive memory swapping by using the Average Page In Alert
Threshold Set parameter (Memory category).










