I'm with a problem about slow backup's using SQLCOMPRESSION. My database runs on a SQL Server 2012 Enterprise edition x64 with 40 CPU's and 256GB RAM.
If i start a backup without compression, my database is like 7.8TB size, and my NIC adapter ( 10Gbit ) reaches like 70% ( between 650mb\s and 750mb\s )
If i use my backup with compression, my database get's like 1.8TB which is very good, but my NIC adapter just reaches 12% ( between 100mb\s and 150mb\s ) maximum, and always transfer in a constant speed, with no up's and downs.
Part of my "batch" file:
This behaviour is normal ? I don't expect with compression to get the same speeds, but 12% of my 10GBIT NIC is very low...
Solved! Go to Solution.
Is your backup job failing to meet its SLA?
See also some Windows type IO thoughts:
Is everything on just few volumes (e.g. less than two or three drive letters)? Or is your OS (1), application binaries (2), database (3), transaction logs (4), and application text logs (5) all assigned/installed to different drive letters? More than five different volumes?
The more I think about this, the less I/we know about your configuration. It's very hard to comment on what changes/tweaks/improvements to make. All I can suggest are some things to check and/or consider.
Is the SQL database on local disk within the server? Or on a SAN? If it's from a SAN, then is the NTFS volume that the database resides upon from one big SAN LUN, or many smaller SAN LUNs? Usually an NTFS volume made from many smaller LUNs will usually result in more possible IO in flight (i.e. more IO queued pending completion), i.e. more disk IO request can be in progress at any one time... but then there are also always limits on the far side of storage - i.e. at the storage array itself. And there's buffer to buffer credit exhaustion to consider inside a SAN fabric. Hmmm, this list is near endless. It all depends how far you are willing to go to investigate all of this?
We haven't identified where the bottleneck is (and there is always at least one bottleneck). Maybe your backup media server (and/or its target storage) cannot ingest any quicker. If your source backup client disk volumes are not 100% busy and have low IO queues (less than 1.0), and your total CPU usage is quiet-ish (but then have you checked for a single max'ed out logical CPU?) and your LAN (backup client to LAN switch) is not maxed out, then this hints at a bottleneck outside of the SQL server - so.. where is it? Are there multipel LAN switches with ISL (inter-switch link) between them? Are they maxed out? The link from switch to media server HIC - is that maxed out? Is the NetBackup media server maxed out? Search this forum for the terms 'waited delayed' for examples on how to spot where delays are occuring.