OpenZiti in time-sensitive environments (Elasticsearch etc.)

I already wrote Philip regarding this as we were in contact initially but thought this might be interesting for others as well:
How does OpenZiti perform in environments where low latency is key?
I am in particular asking because I am considering using OpenZiti in between Elasticsearch nodes and also for communication between my server and the elastic master node.
Communication between the nodes currently has approx .2ms delay.


Every single situation is different. Once the connections are established to the overlay network, in my experience it’s been no different than regular Internet. So it’s subject to how good your Internet link is. Some places have rough Internet, the fabric should help there but it can’t fix everything.

I can tell you as a human, I rarely ( maybe never ) notice that OpenZiti is running. I also don’t know of any great write ups on latency to share. If you’re establishing a single connection and sending data, I think it’s going to be great. If you’re opening and closing connections ( actually tearing them down ) then you’ll have the initial time for any connection in your timing and that certainly could add up, I don’t expect that to be how it works though in a low latency type of app.

Hope that helps. Maybe another from the community has more to share.

1 Like

I have found that correct sizing of the machine to be important… especially if the controller and edge router is running on the same machine… which is very small.

1 Like

Okay, I did some shallow testing:
For my two servers, out of 10 tests, I have a latency between SYN-SYN ACK of 2.8-3.2 milliseconds.
With the connection tunneled over the OpenZiti overlay, I get times for the same tests of 6.2-10.8 milliseconds.
Not too bad so far! I’ll do some testing again to test if there are advantages to latency once the TCP connection has been established.