[NIFI-4834] ConsumeJMS does not scale when given more than 1 thread

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.6.0
• Component/s: Extensions
• Assignee: Mark Payne
• Reporter: Mark Payne
• Created: 31/Jan/18 19:31
• Updated: 19/Jun/20 12:07
• Resolved: 06/Feb/18 20:26

Description

When I run ConsumeJMS against a local broker, the performance is great. However, if I run against a broker that is running remotely with a 75 ms round trip time (i.e., somewhat high latency), then the performance is pretty poor, allowing me to receive only about 30-40 msgs/sec (1-2 MB/sec).

Increasing the number of threads should result in multiple connections to the JMS Broker, which would provide better throughput. However, when I increase the number of Concurrent Tasks to 10, I see 10 consumers but only a single connection being created, so the throughput is no better (in fact it's a bit slower due to added lock contention).

Issue Links

• causes: NIFI-7563 Optimize the usage of JMS sessions and message producers
• links to: GitHub Pull Request #2445