Re:How to Manage Connections Efficiently in Postgres, or Any...
by hinkley on Tuesday October 16, 2018 @06:03PM
(#18233694)
I don’t have the numbers but I do know we are facing this problem at work. When you have microservices that are supposed to respond in <30 ms connection setup time becomes noteworthy.
Ironically we exacerbated the symptoms by reducing average calls per page below 1.0. The burstiness of the traffic made it clear that the low water mark for the connection pool was not working out.
Pooling introduces numerous other issues as well. For example, clients probably shouldn't all be using the same entitlements to connect, right? Surely there are some segregation of duties and roles that clients will be using alternative entitlements into the system?
Finally, you have to show some before and after results. This article simply asserts that connection pooling would improve the situation, but does nothing to demonstrate the point. Even the details of the server matter...how many cpus, how many drives, HDD or flash, what file system is being used?