Backup job from all flash source shows source as bottleneck

[mikegodwin]

I have an all flash VMware vSAN cluster and it backs up pretty fast. The processing rate usually shows 500-800 MB/sec, so I'm not unhappy with that, but the bottleneck always shows Source. I am worried I have something misconfigured somewhere. I have Windows proxy VMs running on each ESXi host. The backup target is a single server with 16x NL-SAS drives in RAID10, so in theory that could allow close to 1GB/sec write throughput to the target at RAID10, but I would still think the bottleneck would not be the source. The VMs in the backup job all show [hotadd]. The vSAN cluster has almost 100 capacity flash drives, so it raises an eyebrow when it shows the source is the bottleneck.

Here is an example of a backup job bottleneck stats:

Source: 79%
Proxy: 25%
Network: 33%
Target: 17%

The only things I can think of changing would be enabling jumbo frames on the network, or changing the vSAN storage policy to use stripe width = 2 (currently it's SW=1). Any thoughts on this?

[Federicof]

What's the connection speed between the source and the proxy? and is there any VMkernel adapters on that network? That will slow it down

[mikegodwin]

25Gbps switching. There are VMkernel adapters for management VLAN and for vSAN VLAN but not on the VM network VLAN.

[foggy]

Hi Mike, please note, that unless we're talking about ideal setup, one of the components in the entire processing chain will still make others to wait and be identified as the bottleneck. In this sense, every environment has the bottleneck and there's always space for improvement, but if you're satisfied with the overall backup performance and meet your backup window, there's nothing to worry about.