How to speed up a Proxy server?

[Wabun]

Hi,

I am using a virtualized Windows 7 pro as a proxy server, and wonder what will provide more speed, the amount of vCPU or Memory.
Just wonder because my memory is a little bit limited on the ESxii server.

Thank you.

[foggy]

Nick, provided the amount of memory meets the proxy server system requirements, the amount of CPU cores is what defines the proxy server performance.

[veremin]

Also, it might be interesting to know whether the proxy is the slowest component in backup chain indeed? Is that what full bottleneck statistics suggest? Thanks.

[Wabun]

Hiya,

Current configuration is 4GB and 2vCPU and the NIC is a E1000.

Yes it is showing as: Bottleneck: Target.

I will change it to 3GB and 4vCPU that should be giving more power reading the link provided?

Thank you

[foggy]

Depends on the backup load you're assigning to this proxy.

[NightBird]

Bottleneck: Target ==> the Backup Repository is the bottleneck, isn't it ??? (or ESXi datastore performance if on replication job)

[Wabun]

Hiya

I am migrating a few VMs, will do a new move tomorrow, will let know the outcome.

Thank you

[foggy]

Are you running backup or replication jobs (or both)? If backup, what kind of repository do you have?

[Vitaliy S.]

I will change it to 3GB and 4vCPU that should be giving more power reading the link provided?

Bottleneck: Target ==> the Backup Repository is the bottleneck, isn't it ??? (or ESXi datastore performance if on replication job)

NightBird is correct, adjusting proxy server configuration will not improve your job performance, you should be looking at target storage hardware here.

[Wabun]

Hi,

I am not backing up, I am migrating, so this means the datastore is the issue?

Kind regards,

[veremin]

Migrating? Using what particular functionality? Quick Migration or Replication?

[Wabun]

Veeam 9, using the Quick Migration Job.

[veremin]

Yep, the destination datastore seems to be the slowest component in this case. Thanks.

[NightBird]

Check if you have write back cache on the destination esx raid controller

[Wabun]

Hiya,

Yes it's an LSI 8708eEM2 with BBU and write back cache enabled.

[dellock6]

also, remember that bottleneck stats are not meant to warn you of any error or problem, it's more an analysis of the time consumed by each component of the process. If the target is the bottleneck, it may simply be that the destination datastore is not able to go any faster when receiving data of Quick Migration.

[Wabun]

Thank you

[foggy]

Check the transport mode used to populate the target datastore.

[Wabun]

Check the transport mode used to populate the target datastore.

Network mode.

It seems that I don't understand the information regards the Bottleneck.
After testing this weekend I discovered that the bottleneck might be the 100Mb nic port, I moved a 120GB VM across the Internet and the limits where the ones my bandwidth graph in DC showed. I am not sure if any proxy server can speed that up... I am worried now when to move a 300GB VM. I also noticed that the amount of data read and transferred was different, Processed: 120GB, Read: 120GB, Transferred: 71.8GB. Does that mean that only actual data is transferred and not deleted files on the disk, I read somewhere that zero filled sectors where not transferred but a bit puzzled about the difference as I can't imagine that there is any zero filled sector left on that disk

Would a proxy server on the Target definitely speed this up?

Kind regards

[foggy]

Having proxy server close to the target datastore will definitely speed things up as traffic will be transferred between proxies in a compressed format (compression is, btw, the reason of why you see Transferred less than Read). Also, it will allow to use hotadd to populate the target datastore, instead of network mode.

What is your current Quick Migration topology?

[Wabun]

What is your current Quick Migration topology?

Veeam Backup and replication 9

Esxi v5.x with a Win7 Backup proxy server, 4vCPU, 3GB, 32GB, E1000
Esxi v5.x with a Win 10 Backup proxy server, 4vCPU, 3GB, 40GB, E1000

When using the proxy the speed is increased from 8MB to 30MB, but now Bottleneck is Proxy.
Source 0% > Proxy 16% > Network 0% > Target 12%
In vSphere Client I can see that each Backup proxy is now having attached the disk from the VM which is being migrated, so it works like you said

Is it possible to get more speed, if so how?
Is there any relation with the size of the Backup Proxy server HD, or is it simple just adding more vCPU?

Thank you.

[foggy]

Just adding more vCPU.

[Wabun]

Thank you.

O, one question if I change vCPU from 4 to 8 do I need to rescan or update this in the Veeam Master?
Or will it automatic detect the new amount of vCPU?

[veremin]

Increase max concurrent tasks from 4 to 8 in settings of given proxy server. Thanks.

[Wabun]

Thanks again, I did some testing with the 8vCPU, but it's back to 8MB speed 'weirdo', I just wonder if it is very important where the Master is located?

- Would you use the Master on the receiving side or on the sending side?

[foggy]

Provided everything else stayed similar to the previous test, just adding vCPU shouldn't result in performance decrease. Though, it doesn't guarantee faster processing as well unless proxy CPU was maxed out previously.

[Wabun]

My fault, I just discovered I had to go in the Master, then to the proxies and then change the cpu to 8 and fo rthe master as well...

Sadly, once the job is running you need to let it run out completely, so I need to wait patiently till this job is done and try again.
Well, it's all about learning

Thanks everyone so far for all feedback, it is much appreciated.