Veeam Proxy and Wan Accelerator

[zbouby]

Hi all !

We have a 2 sites vSphere 6 environment with a 1G link between them.
On the primary site we have one physical server (Dell R730XD, 32GB RAM and E5-2630 V3 CPU,80TB DAS) and Veeam 9.5 is running on it.

This Veeam manage the backups (around 16TB) of around 100 Vms.

Our Veeam make replicas from the primary site to the second one but for now, theses jobs are using bandwidth and we wants to optimize that.
That's why i'm looking at Veeam Proxy + Wan Accelerator.

And so i'm thinking about doing the following.deploy 2 physical servers on each site (Primary + Secondary) with Veeam Proxy + Wan Accelerator.

The 2 physical servers will have the following specs:
(1 x Xeon E5-2623V4 - 2 x 16 Go - 1 x Smart Array Card P840ar_2GB FBWC 12Gb - 1 x SSD 120GB - 2 x 300 Go SAS 15K - 12 x 4 To SATA 7,2 T - 1 x NIC 10G 2 Ethernet Ports - 2 x Power DC 500W - ILO

someone told me that i need to connect a NAS to each server for the Dumps DBS but i'm not sure about it..

So i would like to know and if someone have already implement it and if the previous specs can correspond to my infrastructure.

Many thanks at all for your help/feedbacks.

[DGrinev]

Hi zbouby and welcome to the community!

WAN accelerator recommended in situations with a slow WAN connection (less then 100mb\s).
What exactly do you want to optimize, amount of data transferred (increased transfer time expected by using WAN accelerator) to the remote site or reduce the workload on bandwidth?

someone told me that i need to connect a NAS to each server for the Dumps DBS but i'm not sure about it..

Could you clarify the situation around NAS, and what are those Dumps DBS?

By the way review existing topic about best practices for Replication.

Thanks!

[zbouby]

Hi,

Thanks !

Indeed, i just read the following KB : https://www.veeam.com/kb1877... and as we have a 1G link between our 2 sites, WAN accelerator will have no effect..
And yes, the aim is to have replication between our 2 sites with a 2 hours delay (our RPO is about 4h).

I was thinking about using Veeam Proxy on phyiscal servers on each sites.

But do you think this can help to reduce the time for replicating the primary site to the second one in the aim to have a DR plan ?

Regarding the NAS i think it was about the caching for the Wan Accelerators...but no matter..

[DGrinev]

Hi Pravin,

The primary reason for physical proxies is to get Direct SAN/NFS, so it doesn’t reduce replication time as your WAN connection going to be a bottleneck.

What’s your current bottleneck stats for the replication job? Depending on the reported bottleneck you will understand which component is the "weakest" for this configuration.

Thanks!

[zbouby]

Hi,

Sorry for the delay:
For the replication : Bottleneck is Target with a processing rate of 15MB/s
For backups : It's the source

[DGrinev]

Hi,

Bottleneck target means that the write speed to the target storage is the weakest point in data processing chain. I would recommend you to setup backup proxy VM on the target site to be able to utilize hotadd mode for writing data. This should help to improve performance on the target.

Which backup mode are you using for the backup jobs?

Please, review FAQ topic as it could be useful for further optimizations - bottleneck analysis.
Also, see this existing thread about Replication target host selection.

Thanks!

[zbouby]

Wes i'll have a look at this i already deployed a VM for Veeam proxy at source but now i need to deploy a new one on target site, hope that it'll help !

[zbouby]

Hi,

I deployed 2 VM Veeam proxies, one on source one other on target but i'm not able to have more than 10mbps on the 1G link between our 2 sites.
I'll plan a End to End test to confirm that the link is up and running and if that's the case, i think the issue is located on veeam.

anybody have the same issue or some ideas to use more speed ?

[DGrinev]

Zbouby,

Did you choose the source and the target proxy in the replication job on the data transfer tab?
What was the new reported bottleneck during this latest job run?

Thanks!

[zbouby]

Yes i configured it and for the first time i was able to have 600Mbps for trafficm for the second time 15Mbps, do you know why ?

The traffic is configured for LAN target (1Gbps) and 2 virtual proxies are installed on source/target.

[DGrinev]

Zbouby,

Can you please provide us bottleneck stats reported for both job runs?
As, unfortunately i cannot see it from your post. It can be found in the backup console, job session log.

Thanks!

[zbouby]

Hi,

the bootleneck is the source, but, i changed the job settings for LAN target to WAN target and now it seems that i'm having 800MB/s... strange things, i'll have a look at this to validate that it's the case every replication cycle.

And just another question...how to reduce/optimize the time of the replication job ?
Because for now, it's around 30mns for 2 Vms with a total of 600GB...
as i have around 100Vms to be replicated and with a RPO of 4h, do you have any tips/tricks to achieve that ?

Many thanks for your help

[foggy]

If bottleneck for replication job is target, then please check the transport mode used to populate target datastore with replica VM data. Seems hotadd proxy is not utilized on target.

[Vitaliy S.]

Additionally, you can review what operation takes most of the time in the job session log and then decide how it can be improved further.

[zbouby]

Hmm it seems that when CBT is not used, the rate is more faster [hotadd] are read at around 500- 900MB/s...


When CBT is used, hard disk [hotadd] are read at around 3MB/s...

Becoming crazy

[Vitaliy S.]

What's the difference in the total job time? Read rates can be large due to skipping zeroed blocks in the VMDK.