VMware Cloud Community
zeebahi
Enthusiast
Enthusiast
Jump to solution

vSAN across WAN

Hi everybody,

I am trying to find if there is any maximum  network latency limit for vSAN.

We have to sites  in two different continents, the round trip time between the two is 230msec. We are wondering if we can still do vSAN with cluster stretched between these two locations.

Thanks and have a good night!!

0 Kudos
1 Solution

Accepted Solutions
depping
Leadership
Leadership
Jump to solution

No, this is not possible. This would severely hurt your application performance, to the point where you will have all users of the application shouting at you. If you want to increase availability, consider using async replication and SRM for orchestration. This will provide you the ability to failover VMs after a failure between the locations.

 

Keep in mind, a Stretched Cluster means ALL writes need to be acknowledged by BOTH locations before they are acknowledged to the VM, in other words, every write IO will take 200+ MS. I am sure that would not be acceptable in terms of performance.

View solution in original post

3 Replies
daphnissov
Immortal
Immortal
Jump to solution

Yeah that's not going to work at all. vSAN does have a max latency requirement and that's way too far.

0 Kudos
zeebahi
Enthusiast
Enthusiast
Jump to solution

Thanks  !!!

I just found out:

Network latency
  • Maximum of 1 ms RTT for standard (non-stretched) vSAN clusters between all hosts in the cluster
  • Maximum of 5 ms RTT between the two main sites for stretched clusters
  • Maximum of 200 ms RTT from a main site to the vSAN witness host
0 Kudos
depping
Leadership
Leadership
Jump to solution

No, this is not possible. This would severely hurt your application performance, to the point where you will have all users of the application shouting at you. If you want to increase availability, consider using async replication and SRM for orchestration. This will provide you the ability to failover VMs after a failure between the locations.

 

Keep in mind, a Stretched Cluster means ALL writes need to be acknowledged by BOTH locations before they are acknowledged to the VM, in other words, every write IO will take 200+ MS. I am sure that would not be acceptable in terms of performance.