I have seen sequential READ throughput as low as 8MB/sec on a fibre-connected NetApp. (sorry I don't remember which model it was). With your setup, there could be several different problems. On the format, you are writing about 17MB/sec, but that's going over the network once, from your TSM server to the NetApp. On the client backup, you are going over the network twice - from the client to the TSM server, then from the TSM server to the NetApp. So you could have network issues on either/both legs, AND issues with NetApp performance. My understanding is that you can do some tuning in the way you lay out your RAID sets on the NetApp - you want to get as many disks spinning at once as possible to improve performance. But you can also outrun the ability of the NetApp to purge its cache. You may have to do a LOT of testing and configuring with the NetApp to get improvements. For your testing, try using a TSM DB backup or an EXPORT, so that you are going over the network only once for your comparison tests. You might want to open an incident with your NetApp support to get their ideas of the best way to optimize throughput. From: ADSM: Dist Stor Manager on behalf of Sam Sheppard Sent: Wed 9/20/2006 4:40 PM To: [email protected] Subject: STGPOOL on Netapp
Anyone tried putting a disk storage pool on an ISCSI, GigE attached Netapp-960 box? If so, any recommendations on TSM performance? We are attempting this as a temporary solution, but are having VERY poor performance. Formatting a 250GB volume has taken up to 4 hours and then the backup performance on a test from 1 small client is only pushing 3-4 GB/hour. Also tried a device class of FILE and didn't see much difference. By contrast, equivalent test on a local disk takes just over 1 hour to format and throughput is 27GB/hour. Server is: Solaris 10, TSM 5.3.3. sun fire v240 2 1.5 GHz CPU's 8GB memory Thanks, Sam Sheppard San Diego Data Processing Corp. (858)-581-9668
