[HDFS-2243] DataXceiver per accept seems to be a bottleneck in HBase/YCSB test

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Won't Fix
• Affects Version/s: 0.23.0
• Component/s: datanode

Environment

Using Fedora 14 on a quad core phenom system

• Assignee: Unassigned
• Reporter: Eric Caspole
• Created: 10/Aug/11 19:37
• Updated: 28/Sep/15 21:08
• Resolved: 26/Sep/11 13:37

Description

I am running the YCSB benchmark against HBase, sometimes against a single node, sometimes against a cluster of 6 systems. As the load increases into thousands of TPS, especially on the single node, I can see that the datanode runs very high system time and seems to be bottlenecked by how fast it can create the threads to handle the new connections in DataXceiverServer.run. By "perf top" I can see the process spends about 12% of all its time in pthread_create, and in hprof profiles I can see there are tens of thousands of threads created in just a few minutes of test execution.

Does anyone else observe this bottleneck? Is there a major challenge to using a thread pool of DataXceivers in this situation?

Issue Links

• relates to: HDFS-918 Use single Selector and small thread pool to replace many instances of BlockSender for reads