Veeam is slow with CBT 10-30MB/s

[righter]

Hi

we're using now Veeam since some weeks on a brand new infrastructure.

- Everything is connected with 2x10Gig network
- ESX Servers and Veeam Servers are not having heavy load on memory or cpu
- We have a new Netapp Metrocluster with SSD caching (Using NFS)
- We are using 4 backup proxies, and everyone of them are handling 1 Thread
- Backup Mode is incremental
- Storage Options in Job: Dedup on, exclude swap, compression: optimal, Storage optimization: local target
- CBT is on
- use storage snapshot is on
- Backup Storage Destination is another Netapp Cluster
- Version is 8.0.0.2021

So the problem is that the Read is most of the time between 10-30MB/s.
And Veeam marks always the source as bottleneck. This new storage should be much faster.

Here an example:
16.06.2015 00:52:16 :: Festplatte 2 (50,0 GB) 1,9 GB read at 17 MB/s [CBT]
16.06.2015 00:54:46 :: Busy: Source 99% > Proxy 15% > Network 0% > Target 0%

I have no idea where to search.
The network is definitely not the problem we have an amazing speed.

anyone an idea?

[PTide]

Hi righter,

Do you mind if I ask you something? How is your write speed, is it ok or it's also slower than you expected?

Thanks.

[foggy]

What transport mode is being used during backup (look for the mode [tag] after the assigned proxy server name in the job session log)? Have you configured everything right to utilize storage snapshots on NetApp?

[righter]

@PTide Do you mean the normal write speed from the VM's to the Storage? That speed is great

@Foggy We can only use Network Mode, that setting is on automatic mode (storage snapshot) on our Netapp:
17.06.2015 05:41:04 :: Using backup proxy xxxx for retrieving Festplatte 2 data from storage snapshot

Yes we have double checked, and everything is configured correctly

[foggy]

I suggest to let our engineers take a look at your environment.

[PTide]

Hi righter,

Correct me if I' got it wrong - the same storage, but reads are awfully slow while writes are ok?

If correct, then, I think, I have a guess...

You've mentioned Netapp Metrocluster. As I know there is a mode for NetApp Metrocluster such that writes go to both storages (local and remote) while reads are processed on local one. So my question is: are you sure that your reads are performed on your on-site storage?

Thank you.

[righter]

@foggy ok thanks, I'll check that with my colleagues

@PTide. Hmm we haven't tested write vom Veeam to our Storage. But the storage which Veeam reads from it is the same as the ESX are running on it, and they have excellent speed R+W.
What guess have you in your mind?

[righter]

hey we found the main problem.
Our Repo Server was running out of his 24Gig RAM. we increased the RAM and now it's workin much faster.
Before there were a lot of connections drop until the Repo server was crashed...