On 11/3/06, Mark Perry <[EMAIL PROTECTED]> wrote:
An OS that need to cache "active" working sets, only needs to do so when it runs out of sufficient Central storage. By moving Central storage to Expanded storage one only exacerbates the problem.
I'm not sure you draw the right conclusion, but that may be because I'm not certain what it is ;-) Expanded storage is time stamped and pages selected LRU (which is considered the best we can to predict the near future). Pages in main memory are not selected LRU but WAG. Because we have expanded storage the penalty for WAG vs LRU is minimal when we pick the wrong pages. Expanded storage should be so large that it allows the virtual machine to discover the mistake and get the page back before it is migrated out. The experiment is fairly simple if you do the measurements. Last time I tried this in a memory constrained configuration, taking some memory out and use it as expanded storage lowered my external paging rate a lot and made my transactions 30% faster. YMMV. Only when you have so much real memory that don't do *any* paging to DASD ever and you will not do in the future when your virtual machines grow, then you could in theory reduce a small amount of CPU cycles by trading your expanded storage against real memory. Rob -- Rob van der Heij Velocity Software, Inc http://velocitysoftware.com/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
