The limit of 100GB isn't cast in stone - it's a rule of thumb that I've heard tossed about frequently. The actual fall-off in performance will depend a lot on your disk subsystem. Faster RPM drives can help, as well as spreading the I/O across more spindles (i.e., striping), use of write-cache, etc.
When you reach this limit, the performance degradation can happen fairly quickly. Keep your eye on expiration. If you can't expire objects out of your database in the time you have allotted to this, then your database size can start ballooning, causing a negative feedback loop. Performance will continue to deteriorate quickly from this point. We took the road of getting bigger and faster servers for years, before we finally fell off the cliff. Keep in mind that you don't have to necessarily add another physical server. On AIX, we run multiple TSM server instances, all on the same AIX box, primarily to keep the database sizes down. Another factor to consider, besides performance, is recoverability. If you ever have to restore or audit your database, a larger database can be a serious problem. ..Paul At 04:16 AM 9/1/2005, Hans Christian Riksheim wrote:
Hello, we are experiencing throughput problems on our TSM-installation. Apart from he obvious that we have too few tape drives, bottlenecks in our LAN and an old AIX-box, we got a suggestion to add another TSM-server. The reason was that the performance degrades when the TSM-database reaches about 100Gb in size. We find it a little troublesome to add a new TSM server every time we reach 100 Gb, we would rather just buy us a new AIX-box which is 5 times faster. So I ask you if this 100 Gb limitiation really applies. What is your experience with this? Server version 5.2.3.3 Ca. 100 clients. Best regards Hans Chr. Riksheim
-- Paul Zarnowski Ph: 607-255-4757 Manager, Storage Systems Fx: 607-255-8521 719 Rhodes Hall, Ithaca, NY 14853-3801 Em: [EMAIL PROTECTED]
