Laurent Bendavid <[EMAIL PROTECTED]> wrote: >Stapleton, Mark wrote: > >>From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED] On >>Behalf Of Egon Blouder >> >> >>>I am running a few huge file servers (~6-9 million files, >>>600GB-1.2TB data, 2GB MEM). In order to enhance backup speed I >>>tried to set resourceutilization to a higher value (5-10). >>>After setting that value I monitored dsm process during >>>backup. The process uses up to 2GB memory and exits with a >>>memory error message. >>>TSM Client ran out of memory. >>>Now I'm thinking about installing up to 8GB (max. for W2K Adv. >>>Server) additional memory for these clients. I'll set /PAE >>>boot option in boot.ini in order to enable more than 4GB of memory. >>> >>>Unfortunately I cannot figure out whether TSM client (5.2/5.3) >>>is able to use that additional memory using Address Window >>>Extension (AWE) API. Does anybody know whether it's supported? >>> >>> >> >>If using PAE doesn't get you what you want, try this workaround: >> >>1. Create multiple dsm.opt option files, each using a different nodename >>and pointing at distinct disk drives. Example: >> >> nodename nodeA >> domain c: d: e: >> >> nodename nodeB >> domain f: g: systemobjects >> >>2. Create a TSM scheduler service for each nodename. >>3. Run backups (and restores) in parallel. >> >>-- >>Mark Stapleton ([EMAIL PROTECTED]) >>Office 262.521.5627 >> >> >> >> >We have the same problem. > >If you use TSM Win32 bits, you could not use more than 2GB memory, after >dsmc exit with "ran out of memory". >/PAE didn't work in our tests. > >You could use only one dsmc process for one drive, so you have to check >if you don't exceed this limit for one drive >(depends on the number of objects inspected, 4 millions in my situation). > >However, it's this requisite is ok, the workaround explain is ok. > >We are in the bad case, so we use other workaround : >- split backup with // dsmc on different file tree >- we realize that dsmc in command line have less memory consumption than >dsmc sched > (we have a PMR open ont this object) >- you could use memoryefficientbackup = yes => impact on backup time >
Recently I read a Field Guide (or something like that) on how to run multiple schedulers at one time and how to feed them into one node. I know that I could use memoryefficientbackup=yes but my box has that much memory so why couldn't it be used? Additionally I disklike the implementation of memoryefficientbackup parameter. I would prefer setting the max memory limit for dsm processes. Why don't dsm processes use those 2GB memory and if more memory is needed to process backup dsm should move on memoryefficient behaviour. Btw.: Can anybody forward that to TSM Client R&D? Regards -- __________________________________________________________________ Switch to Netscape Internet Service. As low as $9.95 a month -- Sign up today at http://isp.netscape.com/register Netscape. Just the Net You Need. New! Netscape Toolbar for Internet Explorer Search from anywhere on the Web and block those annoying pop-ups. Download now at http://channels.netscape.com/ns/search/install.jsp
