I have been testing out the WAN accelerator and have strange results. I have two sites connected with a 150 meg WAN that backup each other and are a DR site for each other. Running a backup copy job without a WAN accelerator I get a transfer of around 16 MB/s which is what I would expect because we encrypt our wan and other traffic. A single virgin job though the WAN accelerators will top out at 11 MB/S. I checked the CPU, Memory, Disk of all the pieces and nothing is at 100 percent usage. The source WAN has a cpu at 68 percent but has 7 more cores. I created 4 more WAN accelerators and then I can push over 38 MB/s though, which is more than the 17.88 MB/s that is possible, so I know it is using cache. Each of the four jobs seems to cap out at around 11 meg. I then reran a single job and again and it was still around 11 MB/s. The 11 MB/s is what I would expect with a 100 meg connection so I checked the whole network path and have 1gig the whole way. The increase in bandwidth I get by using 4 is great but why with one job is it such a big hit when nothing is maxed. I tried increasing CPUs,memory, better raid for WAN cache and both physical server and virtual servers for all the parts but results were the same. I know the book says for WAN links under 100 Mbps but is there a hard lock at 100? Anything I might be over looking? By running the job without the WAN accelerator I verified all the parts can handle 16 MB/s.
Again I am happy with the WAN accelerator when I push it . Our plan is to have seven WAN accelerators one mapped to each repository. The hardware is just sitting idle at night so why not. Using with 4 WAN accelerators I already reduced my WAN backup Window so I don’t need to buy more bandwidth. I have been sending live data for days and I should get better than 2 to1 WAN cache.
Sorry for the long post I just want to supply enough information for you to think about my tests.
As you stated, Veeam WAN accelerators are designed primarily for customers with bandwidth <100Mbps, or for customers with faster links that wish to reduce bandwidth usage. You can of course have have multiple WAN accelerators. One thing you didn't post was the bottleneck statistics, which I'd guess is "Target WAN". Once you verify that I can go into more detail, but the primary bottleneck for WAN accelerator is disk I/O performance on the target global cache and the repository (because we retrieve data from the repo if it's not in the cache). Let me know the bottleneck stats and we can dig a little deeper.
. Our plan is to have seven WAN accelerators one mapped to each repository. The hardware is just sitting idle at night so why not. Using with 4 WAN accelerators I already reduced my WAN backup Window so I don’t need to buy more bandwidth. I have been sending live data for days and I should get better than 2 to1 WAN cache.