Details
-
Task
-
Status: Complete
-
High
-
Resolution: Done
-
None
-
None
-
None
-
Ev 18.21
-
Unset
Description
As of now we limit ZMQ message queues, but this limit is per-connection (there is a queue for every connection)
Acceptance criteria:
- create tests (can be without Plenum, just isolated one based on ZMQ) emulating SDK client load more than server can process
- for example 40 requests per sec from client, and 20 reads per sec by listener
- client need to create a new connection for every 5 requests
- create tests for Python and C implementation of ZMQ
- compare memory consumption
- compare Python and C
- analyse how ZMQ deals with a lot of connection (how it reads requests, and releases queues for each connections)
- check if and when memory is released once load is stopped.
- check how Monitor affect memory consumption
Compare memory usage in each case