Parallel Processing and Proxy Appliance

[captainflannel]

With the addition of Parallel Processing, is there is any advantage to having multiple VMProxy appliances? For example in my environment I currently have 1 Proxy appliance in my production, and run HOTADD backups. I have given this appliance 8vCPU to allow for parallel processing. Has anyone gone a different route, like having 3 Proxy appliances with 2vCPU, but still running 3 Parallels jobs?

Other backup vendors use the multiple proxy approach, curious if anyone out there has tried such an option.

[Gostev]

Well, even from redundancy perspective alone, multiple proxies are preferred...

[dellock6]

I'll go with Anton, 1 single proxy is a risk since if it crashes, you loose ALL your proxies. I'd prefer to have multiple smaller proxies than a huge one.

Luca.

[CaptainFred]

I understand the point of having multiple proxies just in case one crashes and to spread load but if the backup server crashes surely it wouldn't matter how many proxies you have because the jobs wouldn't be running anyway?

Also, does this only really apply when running hotadd backups? I run SAN mode backups on my physical backup server - I suppose if I had another physical server that could also have SAN access but then the VBR server would need to transfer the data between the two servers and so they would need fast links between them (my SANs and all connections to them are 10GBE but normal LAN traffic is 1GBE).

[foggy]

I understand the point of having multiple proxies just in case one crashes and to spread load but if the backup server crashes surely it wouldn't matter how many proxies you have because the jobs wouldn't be running anyway?

Yes, so your backup server is virtual, it is always best to protect it with VMware HA or VMware FT.

Also, does this only really apply when running hotadd backups? I run SAN mode backups on my physical backup server - I suppose if I had another physical server that could also have SAN access but then the VBR server would need to transfer the data between the two servers and so they would need fast links between them (my SANs and all connections to them are 10GBE but normal LAN traffic is 1GBE).

It's not quite clear what physical server do you mean here. Is it a second physical proxy or a target host used for replication?

[CaptainFred]

Second physical proxy

[Vitaliy S.]

Why would VBR server need to transfer data between two servers? If you're talking about source proxy (used to retrieve data), then it will send data to your repository over whatever connection you have to this repository. In case of replication, the data would go between two proxy servers over the network.

[tsightler]

Probably because his current physical box is both a proxy and repository, which means all proxy->repository data stays local to the box, while a second physical proxy would have to send data across the network to the repository.

[captainflannel]

Which solution should provide faster results?

Option 1 - Single proxy for HOTADD and 6vCPU able to process 3 jobs

Option 2 - 3 proxies for HOTADD and 2vCPU per proxy, able to process 3 jobs

[veremin]

I don’t think that the results would differ dramatically. Though, you can test both scenarios and see which one best answers your requirements. Though, as mentioned above, it’s still advisable to have multiple proxies, instead of single one, in order to guarantee required redundancy. Thanks.

[dellock6]

Yes, so your backup server is virtual, it is always best to protect it with VMware HA or VMware FT.

Here I disagree. By having separated proxy and repository you can effectively transform the Veeam Server in a management-only server, and the best way to protect it is by installing it in a VM. But at the end only HA is a viable solution, since the cpu requirement makes it almost impossible to protect it via FT.
Veeam Server is indeed a SPOF, but maybe a different solution, in highly demanding enviroment, would be to use tools like NeverFail, but I never tests if it can also support Veeam services...

Luca.

[Gostev]

I agree with both of you, and not sure what do you disagree with, when both are saying the same thing

[mongie]

Gostev... On a scale of 1-10, how much do you enjoy reading and replying to these forums?

[Gostev]

10... unfortunately, too much work lately with v7 release, so I can only do forums briefly and at weird times (like right now, 4:15 am my time)

[dellock6]

And the forums realy exploded after v7 release, the amount of new posts is insane at times

Luca.

[dellock6]

I agree with both of you, and not sure what do you disagree with, when both are saying the same thing

I was disagreing only on the FT part, the 1vCPU limit makes impossible to use it to protect the Veeam Server

Luca.

[Gostev]

Not if you:

having separated proxy and repository you can effectively transform the Veeam Server in a management-only server

1 vCPU is plenty for management-only Veeam server.

[dellock6]

ok, to be more specific, it depends where the SQL server is installed, that is the real bottleneck of the veeam server. When we run several jobs all at once the SQL server is under a fair load. We run 60 jobs per night, and all starts at 21:00, so probably this does not apply to smaller environments...

Luca.

[Gostev]

Agree on SQL part, this is purely for management-only server without SQL.