Physical vs Virtual Proxies went it comes to performance?

[ShawnKPERS]

Is there an inherent performance benefit to using one or the other? I ask because I want to know if there is a good justification for provisioning a physical server to act as a proxy if I can also do it in a VM.

I know for direct san backups I will need to give the VM access to the iSCSI connections of the source SAN and since my storage repository is going to also be an iSCSI SAN I will need to give this VM access to it as well. I have not setup direct iSCSI access to a VMware VM before so not sure if there are any caveats but I know you can do it.

But from a performance prospective are there any real negatives using a VM over a physical box is both systems have access to the same amount of RAM, CPU, and 10 gig links?

Just trying to find a way to justify the cost/ maintenance of dealing with adding an additional physical server. If it might gain me an additional 10 megs a seconds, I would say no but if I can see something like 200 megs then yes it would say it would be worth looking at. Please let me know what you think or have experienced.

[foggy]

Given similar resources, there's no performance difference, transport method (direct SAN, hotadd/NBD) is what plays more important role here. But keep in mind that in case of virtual proxies, you are consuming resources from (presumably) the same environment you are protecting, while physical proxies provide dedicated resources.

[ShawnKPERS]

Thanks, that is a big help, I have been using virtual proxies for some time now and am about to gut and redesign our backup system and just wanted to know if there was any real gain by moving back to physical. I am in a situation where the amount of hardware I have on my ESXi hosts is not saturated to the point were running the proxies on them should be any issue.

[dellock6]

I'd say the main difference is more that on a VM you "may" have other vm consuming resources of the host, thus the proxy being limited. But you can setup a resource pool for the proxies and set reservation for both cpu and ram so that you have all that you need. For the network, the same idea applies, there is other VM traffic going over the same uplink, regardless if dedicated vlan are used, but you can solve it a bit for example using NIOC (if your vmware license allows it).
Also, you may want to have some DRS anti-affinity rule to distribute the proxies over multiple ESXi hosts to spread the load over your cluster.

[ShawnKPERS]

@dellock6 great minds, I currently dedicate one proxy per host using DRS anti-affinity. I will look into NIOC as well but I have been running this way for years currently and just though to ask in case there was some inherent advantage to going physical outside of not having to deal with resource sharing given it is dedicated hardware.