A completely different thought (based on the "millions" of files mentioned earlier):
- Do your LTO5 drive support some type of variable speed matching, if so, I'm wondering if the drive is waiting for buffers to become full before writing. What are you NUMBER_DATA_BUFFERS and SIZE_DATA_BUFFERS set to?
- Assuming the "millions" of files are from a Windows environment, have you investigated this:
I have set this option on several systems that have millions if files and it has helped in backup performance dramatically (your millage may vary).
Scott J.
>>> mstoller <nbu-forum AT backupcentral DOT com> 11/19/2013 12:12 PM >>> With a 10x performance difference between backing up to the LTO4 drives and the LTO5 drives, I'd suggest checking the fibre path from the server to the tape drive in the LTO4 library and comparing it to the LTO5 library. You might want to check port errors at the switch.
I have my doubts this is an issue of the data going too slow and bogging down the LTO5 - I typically don't see such a large performance hit in those cases.
Questions - is anything elses backing up to the LTO5 library?
- if so, is the performance similar to this job? - have you try a different job from this server to the LTO5 and 5 libraries and seen similar performance discrepancies? - have the drives in the LTO5 library been updated to current firmware levels?
I don't have definite answer on this one, but here are a few recent issues I've seen that led to poor backup performance when going to different drives (maybe they'll help point you in the right direction?):
1 - I doubt this is the case at your site, but I have seen an instance where traffic from one media server was routed through a different server before going to the tape drives - so, the data went from client A to media server 1 over the LAN , then out the LAN to media server 2 and then across the SAN to the drives - in this case the LAN was the bottleneck. I believe this was a case of picking the wrong storage unit for a job. It's worth taking a look at.
2 - Another thing to look at: I recently ran into a case at a customer site with slow performance to the drives on one library vs. going to a different library. It turns out that the fast library's drives were plugged directly into the switch that the servers were also plugged into, while the slow library's drives were plugged into a switch that was connected to the first switch with a 1Gb/s link - and there were many devices attached to that second switch, all trying to get data through that 1Gb/s link. Just too much competition for limited bandwidth.
Matt
+---------------------------------------------------------------------- |This was sent by matt.stoller AT pobox DOT com via Backup Central. |Forward SPAM to abuse AT backupcentral DOT com.
+----------------------------------------------------------------------
-- Rusty Major ▪ Technical Architect and Team Lead - Backup Infrastructure Engineering ▪ Sungard Availability Services ▪ 757 N. Eldridge Pkwy., Suite 200, Houston, TX 77079 ▪ 281-584-4693 ▪ Cell 713-724-4914 ▪ rusty.major AT sungard DOT com ------------------------------------------------------------------------------
Think before you Print CONFIDENTIALITY: This e-mail (including any attachments) may contain confidential, proprietary and privileged information, and unauthorized disclosure or use is prohibited. If you received this e-mail in error, please notify the sender and delete this e-mail from your system.
