- Posts: 15
- Liked: never
- Joined: Jan 13, 2011 9:56 am
- Full Name: Chris Hibbert
I am experiencing poor backup performance (or at the very least what I would consider poor performance) throughput using Veeam backup.
Veeam is installed on a 64bit Windows 2008 server, this server has 2 x Dual Core AMD 2212HE (Dual core 2.0GHz Processors) with 32GB of RAM.
2 x iSCSI (Dell MD3000i) attached storage, this is connected through a Cisco 6509 switch, each controller is on a different card and vlan'd off from the rest of the network, the veeam server has network ports allocated into each vlan, and the iSCSI initiator's have been set into failover mode rather than round robin as I have read that this can affect the throughput. The Cisco 6509 has netflow enabled on the relevant iSCSI ports, and netflow rx/tx is enabled on the adapters.
This server connects to the shared storage (CX4-120) thought Brocade Fiber switches running at 4Gbps
We are running 7 x vSphere 4 servers, using a Dell CX4-120 all connected through Brocade switches.
For full backups we are showing speeds of 18MB/s which with the above set up and connection I think is unacceptable, the differential/incremental backups are maxxing out at 23MB/s which also is nowhere near fast enough
I have a DR Site which is running vSphere 4 with a similar MD3000i as storage and accross the WAN I Can achieve speeds of around 50MB/s (over a 1Gb EES Circuit) not sure if this can be improved on much, although I would have hoped for a little more this is using replication.
Steps taken in resolving this
[*]Enabled netflow - this had little to no effect on overall performance
[*]Running to local based drives to check that whether server based issue or network, the local backups seemed to be faster but unfortunately local disk space is a premium and there isn't enough in the server available to test this fully, however I exceeded the full backup speeds just before the backups finished. I am currently creating more local space to test this theory.
I do have access to another server which I "could" switch over which would provide extra grunt (2 x 4core CPU's) if it is thought it'd be the likely issue.
Any help would be great in solving this, I'm sure I must have missed a setting somewhere, but it's getting like I cant see the wood for the trees type thing
Users browsing this forum: No registered users and 28 guests