Is there a communication latency requirement for cluster nodes?

This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: 集群节点有通讯延时时间要求吗

| username: HACK

Does the TiDB cluster have network latency requirements for internal communication between nodes?
For example, in a local environment with poor network performance or remote network communication in a two-site three-center architecture.

| username: xfworld | Original post link

Poor network connectivity will definitely be unsuitable, as there are minimum latency requirements.

| username: 啦啦啦啦啦 | Original post link

As a distributed cluster, TiDB has relatively high requirements for time, especially since PD needs to distribute unique timestamps.

| username: liuis | Original post link

Distributed… definitely has requirements.

| username: Kongdom | Original post link

Getting the TSO from PD is the ultimate truth of the cluster.
:wink: TiDB cannot be without TSO, just like the West cannot be without Jerusalem.

You can refer to the parameters here:

| username: hey-hoho | Original post link

Latency within the same data center is less than 0.5ms.
Latency between multiple data centers in the same city is less than 2ms.
Latency between data centers in different locations is less than 5ms.

This can be used as a reference.

| username: WalterWj | Original post link

Within 3ms, if the duration requirement is high, it needs to be even smaller.

| username: liuis | Original post link

Is this a reference you use in your actual process?

| username: Running | Original post link

The later the delay, the better. Isn’t it all 10 Gigabit Ethernet?