Legacy HDS Forums

Cache management algorithms on USP

Discussion created by Legacy HDS Forums on Jul 27, 2006
Latest reply on Aug 10, 2006 by Legacy HDS Forums

Originally posted by: mackem



Hi,

Can anybody tell me the following in relation cache management on the USP -

What percentage of cache max and min can be assigned to read data?
What percentage of cache max and min can be assigned to write data?
At what point does the USP start destaging data to disk?

I am aware that this is all managed by the subsystem without user invervention but this is very useful information to know when performance testing applications on a USP.

Ive scoured my notes and can find the following for the 9980V which I got from attending the course, but when I check my USP course notes I cant find anything similar -

"The 9980V can assign between 30-70% of cache to either read or write data.
Destaging of data from cache usually starts when cache gets to around 25-30% occupancy.  If it reaches 40% then the system will speed up the de-staging process.  Once over 50% occupancy the system will issue DASD fast-write delay notifications or retry indications to hosts in order to start slowing down inflow of data.  At 70%  occupancy the system goes into controlled panic mode where no ACKs are sent until data is destaged to disk."

Id be keen to know how this is on the USP.

Are there any system option modes that can be set that alter the way that cache is managed?

Many thanks,

mackem

Outcomes