It seems to be one of the topics that don't have an easy answer. On one side, the best cache is the cache that is the closest to the action. Oracle SGA cache is better then the Linux cache. Linux cache is better than MDC cache. MDC cache is better than DASD controller cache. That is when all things are equal. So, in this case, turn off MDC and up the storage in Linux.
However, the storage in Linux is basically dedicated to that machine. When the machine is idle, the storage is still kept for that machine. OK for LPAR, but bad for sharing images in VM. So, in VM, we restrict the amount of real memory for the Linux image, just to the point of active swapping (to a vdisk of course), which decreases the real storage foot print of the Linux image. That leaves more storage for other machines. That tends to be a good thing. In this case you are trading some CPU time for the ability to handle more machines in the same real memory. The real answer will be different for test machines and production machines and what response time is required. We went from escon attached dasd, and MP3000 internal dasd to Ficon2 attached DS6800. What a preformmer! So, for now, we are scaling back our cache use and dumping more I/O to the DS6800. When we start getting unacceptable service times, those machines may get additional storage for their caching. So MDC is good in that it is shared across all machines. But if you have Linux doing effective caching, when Linux needs to do a read, it won't be in MDC anyway. Avoid the CPU overhead and turn off MDC for those machines. Being a VM bigot, I have a hard time dedicating memory to an image (using Linux cache and not using MDC) when I can be using the memory for some other purposes (trim Linux cache and use, perhaps, expand memory for MDC). Tom Duerbusch THD Consulting >>> [EMAIL PROTECTED] 12/9/2005 3:49 PM >>> My VM guy here is saying that he read some place that having mini-disk cache turned on for mini-disk volumes used by linux for system disk is a good thing. My position has always been that Linux is caching what VM is caching and it's a double fault. Since I'm not the VM guru (and neither is he really, been doing VM here for just over a year) I don't have any traction in getting this changed. >From what I can see I'm caching 550 megs in Central stor on average and 163 or so megs to expanded. The only disks I could get him to turn off caching for was my linux swap volumes. So who is right? Isn't it correct that turning off mini-disk caching for the linux volumes is the recommended method? Wouldn't that free up memory othwise used by VM for the caching? We're getting very close to making VM page on a consistent basis. We're running 6 websphere servers in 6 gigs real 1 gig expanded (I know it should be more. We just don't have it). I'm also in a position where I have to recommend how much memory and CPU I think we're going to need over the next three years, but nobody wants to tell me how much more load they want to send me so thats hard to say. Anyway, if theres cold hard fact that can be waved in eithe rmy face or his to settle this, I'd appreciate any input. Have a great weekend! ---------------------------------------------------------------------- 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 ---------------------------------------------------------------------- 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
