[HAMA-932] Use Kryo Serializer for message and bundle serializations.

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.6.4
• Fix Version/s: 0.7.0
• Component/s: bsp core
• Assignee: Edward J. Yoon
• Reporter: Edward J. Yoon
• Created: 10/Mar/15 09:32
• Updated: 24/Mar/15 01:20
• Resolved: 24/Mar/15 01:20

Description

As we already know, kryo serializer shows quite awesome performance. I suggest that we use kryo serializer for everything.

Attached patch uses kryo for message serialization and below is its results:

15/03/10 18:08:08 INFO bsp.BSPJobClient:     TASK_INPUT_RECORDS=30000
15/03/10 18:08:08 INFO bsp.FileInputFormat: Total input paths to process : 3
15/03/10 18:08:09 INFO bsp.BSPJobClient: Running job: job_201503101804_0007
15/03/10 18:08:12 INFO bsp.BSPJobClient: Current supersteps number: 0
15/03/10 18:09:34 INFO bsp.BSPJobClient: Current supersteps number: 1
15/03/10 18:11:24 INFO bsp.BSPJobClient: Current supersteps number: 2
15/03/10 18:11:27 INFO bsp.BSPJobClient: Current supersteps number: 3
15/03/10 18:14:33 INFO bsp.BSPJobClient: Current supersteps number: 4
15/03/10 18:14:36 INFO bsp.BSPJobClient: Current supersteps number: 5
15/03/10 18:17:20 INFO bsp.BSPJobClient: Current supersteps number: 7
15/03/10 18:18:12 INFO bsp.BSPJobClient: Current supersteps number: 9
15/03/10 18:18:15 INFO bsp.BSPJobClient: Current supersteps number: 11
15/03/10 18:18:30 INFO bsp.BSPJobClient: The total number of supersteps: 11
15/03/10 18:18:30 INFO bsp.BSPJobClient: Counters: 11
15/03/10 18:18:30 INFO bsp.BSPJobClient:   org.apache.hama.graph.GraphJobRunner$GraphJobCounter
15/03/10 18:18:30 INFO bsp.BSPJobClient:     ITERATIONS=4
15/03/10 18:18:30 INFO bsp.BSPJobClient:     INPUT_VERTICES=30000
15/03/10 18:18:30 INFO bsp.BSPJobClient:   org.apache.hama.bsp.JobInProgress$JobCounter
15/03/10 18:18:30 INFO bsp.BSPJobClient:     SUPERSTEPS=11
15/03/10 18:18:30 INFO bsp.BSPJobClient:     LAUNCHED_TASKS=3
15/03/10 18:18:30 INFO bsp.BSPJobClient:   org.apache.hama.bsp.BSPPeerImpl$PeerCounter
15/03/10 18:18:30 INFO bsp.BSPJobClient:     SUPERSTEP_SUM=33
15/03/10 18:18:30 INFO bsp.BSPJobClient:     TIME_IN_SYNC_MS=596707
15/03/10 18:18:30 INFO bsp.BSPJobClient:     IO_BYTES_READ=84870046
15/03/10 18:18:30 INFO bsp.BSPJobClient:     TOTAL_MESSAGES_SENT=44717766
15/03/10 18:18:30 INFO bsp.BSPJobClient:     TASK_INPUT_RECORDS=30000
15/03/10 18:18:30 INFO bsp.BSPJobClient:     TOTAL_MESSAGE_BYTES_TRANSFERED=361694719
15/03/10 18:18:30 INFO bsp.BSPJobClient:     TOTAL_MESSAGES_RECEIVED=270035
Job Finished in 639.248 seconds


.... 
Kryo version:

15/03/10 18:23:17 INFO bsp.BSPJobClient:     TASK_INPUT_RECORDS=30000
15/03/10 18:23:17 INFO bsp.FileInputFormat: Total input paths to process : 3
15/03/10 18:23:17 INFO bsp.BSPJobClient: Running job: job_201503101822_0001
15/03/10 18:23:20 INFO bsp.BSPJobClient: Current supersteps number: 0
15/03/10 18:23:35 INFO bsp.BSPJobClient: Current supersteps number: 1
15/03/10 18:24:16 INFO bsp.BSPJobClient: Current supersteps number: 2
15/03/10 18:24:19 INFO bsp.BSPJobClient: Current supersteps number: 3
15/03/10 18:25:23 INFO bsp.BSPJobClient: Current supersteps number: 4
15/03/10 18:25:26 INFO bsp.BSPJobClient: Current supersteps number: 5
15/03/10 18:26:47 INFO bsp.BSPJobClient: Current supersteps number: 7
15/03/10 18:26:53 INFO bsp.BSPJobClient: Current supersteps number: 9
15/03/10 18:26:56 INFO bsp.BSPJobClient: Current supersteps number: 11
15/03/10 18:27:10 INFO bsp.BSPJobClient: The total number of supersteps: 11
15/03/10 18:27:10 INFO bsp.BSPJobClient: Counters: 11
15/03/10 18:27:10 INFO bsp.BSPJobClient:   org.apache.hama.graph.GraphJobRunner$GraphJobCounter
15/03/10 18:27:10 INFO bsp.BSPJobClient:     ITERATIONS=4
15/03/10 18:27:10 INFO bsp.BSPJobClient:     INPUT_VERTICES=30000
15/03/10 18:27:10 INFO bsp.BSPJobClient:   org.apache.hama.bsp.JobInProgress$JobCounter
15/03/10 18:27:10 INFO bsp.BSPJobClient:     SUPERSTEPS=11
15/03/10 18:27:10 INFO bsp.BSPJobClient:     LAUNCHED_TASKS=3
15/03/10 18:27:10 INFO bsp.BSPJobClient:   org.apache.hama.bsp.BSPPeerImpl$PeerCounter
15/03/10 18:27:10 INFO bsp.BSPJobClient:     SUPERSTEP_SUM=33
15/03/10 18:27:10 INFO bsp.BSPJobClient:     TIME_IN_SYNC_MS=185464
15/03/10 18:27:10 INFO bsp.BSPJobClient:     IO_BYTES_READ=84870046
15/03/10 18:27:10 INFO bsp.BSPJobClient:     TOTAL_MESSAGES_SENT=44717766
15/03/10 18:27:10 INFO bsp.BSPJobClient:     TASK_INPUT_RECORDS=30000
15/03/10 18:27:10 INFO bsp.BSPJobClient:     TOTAL_MESSAGE_BYTES_TRANSFERED=573865062
15/03/10 18:27:10 INFO bsp.BSPJobClient:     TOTAL_MESSAGES_RECEIVED=270039
Job Finished in 250.217 seconds