Performance bottleneck tips
Hi TSMers, I need detect the bottleneck of backup performance of TSM W2K client (2xPII 450, 640MB RAM). It is working as fileserver (thousands files with general size about 45GB). many files are expiring, many are rebounded during backup so that means increased server load and lowering of backup performance (I marked it during monitoring of backup). Some other symptoms perhaps suggests problem with network card and switch and possible bad detecting of connection speed (10/100Mbit), but problem can be at TSM client side too. Can I read out something from performance statistics at and of backup (network transfer rate, aggregate rate, data transfer time .. etc.), or from development of server/client processor load during backup? I mean some tips like: If transfer rate is much higher than aggregate rate, problem is on server side ... (that's only example). Simply, how characteristics like TSM server load, connection speed, setting of client communication parameters, rebinding files .. etc. can affect backup statistic values? Any suggestion will be appreciated (drowning man plucks at a straw :-)) ). Many thanks. Tom
Re: Performance bottleneck tips
Hi, I'll do you one example of a node with a network bottleneck, from our act log I took: 04/10/02 14:39:22 ANE4952I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects inspected: 130,542 04/10/02 14:39:22 ANE4954I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects backed up: 891 04/10/02 14:39:22 ANE4958I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects updated: 0 04/10/02 14:39:22 ANE4960I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects rebound: 0 04/10/02 14:39:22 ANE4957I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects deleted: 0 04/10/02 14:39:22 ANE4970I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects expired:796 04/10/02 14:39:22 ANE4959I (Session: 4419, Node: SI_REMCO_PIPC) Total number of objects failed: 0 04/10/02 14:39:22 ANE4961I (Session: 4419, Node: SI_REMCO_PIPC) Total number of bytes transferred:21.55 MB 04/10/02 14:39:22 ANE4963I (Session: 4419, Node: SI_REMCO_PIPC) Data transfer time:1,399.48 sec 04/10/02 14:39:22 ANE4966I (Session: 4419, Node: SI_REMCO_PIPC) Network data transfer rate: 15.76 KB/sec 04/10/02 14:39:22 ANE4967I (Session: 4419, Node: SI_REMCO_PIPC) Aggregate data transfer rate: 11.58 KB/sec 04/10/02 14:39:22 ANE4968I (Session: 4419, Node: SI_REMCO_PIPC) Objects compressed by:0% 04/10/02 14:39:22 ANE4964I (Session: 4419, Node: SI_REMCO_PIPC) Elapsed processing time:00:31:45 As you can see the data transfer time is: 1,399 sec or 23:20 minutes. The remainder of the total processing time (31:45 minutes) is spend by the node examining objects (and this one only has 130k files). Since the time it takes to dl. a fs last state from the server to be able to compare it with the current state of the fs, is small compared to the time it takes to walk the fs, this can be neglected as a factor (even with very little bandwidth, like in this example). If your data transfer time is very small compared to the total time, the bottleneck is in the client node, examining objects... AS you can see, the network transfer rate here is about 15.76 KB/s, which is absolutely unacceptable on a lan, in this case it is something we have to live with... ;) -- Met vriendelijke groeten, Remco Post SARA - Stichting Academisch Rekencentrum Amsterdamhttp://www.sara.nl High Performance Computing Tel. +31 20 592 8008Fax. +31 20 668 3167 I really didn't foresee the Internet. But then, neither did the computer industry. Not that that tells us very much of course - the computer industry didn't even foresee that the century was going to end. -- Douglas Adams
Re: Performance bottleneck tips
Tom, we recently published a viewacct script that tells you quickly whether the problem in the client, the network, or the server. See http://www.servergraph.com/techtip3.htm, Our full-bore product also tells you whether your ENTIRE SITE (not just one node) is suffering most from client slowdowns, network slowdowns, or server delays. Download the demo, and it will probably help you find other problems -- oops, opportunities Hope this helps. - Mr. Lindsay Morris CEO, Servergraph www.servergraph.com 859-253-8000 ofc 425-988-8478 fax -Original Message- From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Toma9 Hrouda Sent: Thursday, April 11, 2002 6:38 AM To: [EMAIL PROTECTED] Subject: Performance bottleneck tips Hi TSMers, I need detect the bottleneck of backup performance of TSM W2K client (2xPII 450, 640MB RAM). It is working as fileserver (thousands files with general size about 45GB). many files are expiring, many are rebounded during backup so that means increased server load and lowering of backup performance (I marked it during monitoring of backup). Some other symptoms perhaps suggests problem with network card and switch and possible bad detecting of connection speed (10/100Mbit), but problem can be at TSM client side too. Can I read out something from performance statistics at and of backup (network transfer rate, aggregate rate, data transfer time .. etc.), or from development of server/client processor load during backup? I mean some tips like: If transfer rate is much higher than aggregate rate, problem is on server side ... (that's only example). Simply, how characteristics like TSM server load, connection speed, setting of client communication parameters, rebinding files .. etc. can affect backup statistic values? Any suggestion will be appreciated (drowning man plucks at a straw :-)) ). Many thanks. Tom