At first, when Mark suggested only -9 and -27 disks, I disagreed with him. After a day, I started agreeing with him. Now I'm back to disagreeing with him. <G> I guess it all depends....
First, you have to go back to hardware. You can only create 256 drives out of any RAID array. On the DS6800/DS8100, if you are using 72 GB drives, if you only define mod-3s, you can't use all the space available. And this gets worse as you go to the 145 GB drives, the 300 GB drives, the 500 GB drives and the 1 TB drives. You just have to have some/many large 3390s in order to use all the space. Mark is pretty much correct, that 3 3390-3 vs 1 3390-9 are the same. If they are on the same raid array there isn't much performance improvement unless you consider that you can only do 1 I/O per device at the same time, unless you use PAV (and paid for that feature). Consider that a mod-27 will be only on a single RAID array. Divide it into 3 3390-9 and you can put each one on a different array. Now you can truely do multiple I/Os at the same time. In my case, I do have some larger Linux guests, but I do have a bunch of small ones. Really they could be on a 3390-1, but, eventually, some of them, due to software installs, tend to grow and I need more space. Their data, however, I either have as LVM (yep, larger capacity drives work well here), or NFS mounted space. So my root drive is usually a 3390-3. HOME and other data directories are on LVM or NFS space. For example, my Oracle servers use LVM for local Oracle tablespace. However, it backs up to a NFS server. That NFS server is an easy place for me to backup all the data from multiple machines to tape. The Oracle server is physically backed up weekly with FlashCopy and that flashed image is then backed up to tape as a physical image. This is used for disaster recovery purposes. My point is that small drives have their purpose. But I also would tie together a dozen mod-3 drives if I had the option to tie together a lessor amount of mod-9 drives (or higher capacity). If you need I/O performance, make sure you spread your data across several arrays. This may require you to do smaller drives. I can sure tell the difference when I copy a GB file when from/to are on the same array vs from/to being on different arrays. BTW, I'm not sure why someone would do this, but you can create mod-10, mod-11 mod-xx drives in the IBM Dasd Subsystems. Any size is really available. But you may face issues of "standards". Just another opinion Tom Duerbusch THD Consulting >>> Ron Foster at Baldor-IS <rfos...@baldor.com> 1/13/2011 1:41 PM >>> Hello list, This may have been discussed before... Way back in deep dark ancient history, we used the Redbook to get started with Linux under z/VM. As a result, we carved up our storage subsystem in to a bunch of mod3 drives. We put a few mod 9 drives in the mix. We added drives to a guest in standard chunks. That is when storage was needed by a Linux system, we added a mod9 or mod3 to it. When that Shark went off lease and we moved to a DS8000, we pretty well kept the same philosophy. Only we added a bunch more mod3 and mod 9 drives. We are a SAP shop and any large databases reside in DB2 on z/OS. There are a few large file systems on 3 or 4 of our Linux systems, but for the most part the drives attached to a Linux system go something like this. A boot drive. One to several mod3 drives for swapping (the appropriate ones have vdisks). One to several mod3 or mod9 drives for the SAP code and local files. We are moving our production drives. We finally have gotten our production Linux systems where about half or do very little swapping. We do not have dirmaint, so we keep up with disk allocations with dirmaint and a spreadsheet. Now time has come to migrate to another storage system. I was wondering what other folks do. 1. Do they have a whole bunch of mod9 and mod3 drives that they allocate to their guests? 2. Do they take mod27 drives (someone at SHARE warned me about taking mod54 drives) and use mdisk to carve them up into something smaller. Any input would be appreciated. Thanks, Ron Foster ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
