Describe the suggested improvement
Is your improvement related to a problem? Please describe.
Currently every queue/topic put uses MQPMO_SYNCPOINT for transaction modes.
IBM MQ does an fsync on EVERY put, this puts a lot or pressure on IO and the storage cannot batch multiple writes limiting throughput
This limits throughput during testing to roughly ~1.500msg per second.
Describe the suggested solution
By having put's done async, and block these for a number of puts (100+) or time (like 50ms) and then do a single Commit.
A quick spike with this approach resulted in significant send throughput increase. (+5x)
Additional Context
No response
Describe the suggested improvement
Is your improvement related to a problem? Please describe.
Currently every queue/topic put uses MQPMO_SYNCPOINT for transaction modes.
IBM MQ does an fsync on EVERY put, this puts a lot or pressure on IO and the storage cannot batch multiple writes limiting throughput
This limits throughput during testing to roughly ~1.500msg per second.
Describe the suggested solution
By having put's done async, and block these for a number of puts (100+) or time (like 50ms) and then do a single Commit.
A quick spike with this approach resulted in significant send throughput increase. (+5x)
Additional Context
No response