Streaming endpoint performance for a python client Vs golang client

[anand kumar]

Hi,

We have a grpc server written in golang and one of the endpoints it serve is a streaming endpoint. For one call, the streaming endpoint returns 100k+ objects where object is a dictionary with all text keys and values.

From a python client, getting all the data takes 40 sec, only 2-3 of which is spent by service executing the business logic and the remaining is in sending all the data back to client. 

When tried from a golang client, all the data is gettable in 3-4 seconds.

So this could be due to some configuration difference in setting up a grpc client in python vs golang, but we have been unable to find any. 

Any help would be appreciated.

[Craig Tiller]

What does CPU utilization look like during those 40 seconds? Are you sitting idle waiting for data, or spinning the CPU doing work?

--
You received this message because you are subscribed to the Google Groups "grpc.io" group.
To unsubscribe from this group and stop receiving emails from it, send an email to grpc-io+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/grpc-io/144f6fb5-dd50-47d1-b569-70269ca2e65an%40googlegroups.com.

[anand kumar]

It's the latter. top shows CPU at 100+% for the entirety of 40s for the client process.

[anand kumar]

fwiw, got another similar datapoint with an older service of ours serving a streaming endpoint and written in golang. The time taken by python client in getting all the objects is more than double of what the golang client is spending.