[
https://issues.apache.org/jira/browse/HDFS-9710?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tsz Wo Nicholas Sze updated HDFS-9710:
--------------------------------------
Attachment: h9710_20160224.patch
h9710_20160224.patch: adds a test.
The new test shows that #IBR is greater decreased from ~550 calls to 41 calls.
- Batch mode IBR
{code}
2016-02-24 18:46:04,197 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:runIbrTest(134)) - batchModeIbr=true, duration=6.487ms,
createFileTime=68.188ms, verifyFileTime=35.348ms
2016-02-24 18:46:04,351 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:57961:
IncrementalBlockReportsNumOps=41
2016-02-24 18:46:04,362 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:57965:
IncrementalBlockReportsNumOps=41
2016-02-24 18:46:04,374 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:57970:
IncrementalBlockReportsNumOps=41
2016-02-24 18:46:04,386 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:57974:
IncrementalBlockReportsNumOps=41
{code}
- Immediate IBR
{code}
2016-02-24 18:46:10,748 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:runIbrTest(134)) - batchModeIbr=false, duration=4.678ms,
createFileTime=67.169ms, verifyFileTime=7.548ms
2016-02-24 18:46:10,756 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:61066:
IncrementalBlockReportsNumOps=553
2016-02-24 18:46:10,764 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:61071:
IncrementalBlockReportsNumOps=561
2016-02-24 18:46:10,772 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:61075:
IncrementalBlockReportsNumOps=558
2016-02-24 18:46:10,780 [main] INFO datanode.TestBatchIbr
(TestBatchIbr.java:logIbrCounts(154)) - 127.0.0.1:61079:
IncrementalBlockReportsNumOps=558
{code}
> Change DN to send block receipt IBRs in batches
> -----------------------------------------------
>
> Key: HDFS-9710
> URL: https://issues.apache.org/jira/browse/HDFS-9710
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode
> Reporter: Tsz Wo Nicholas Sze
> Assignee: Tsz Wo Nicholas Sze
> Attachments: h9710_20160201.patch, h9710_20160205.patch,
> h9710_20160216.patch, h9710_20160216b.patch, h9710_20160217.patch,
> h9710_20160219.patch, h9710_20160224.patch
>
>
> When a DN has received a block, it immediately sends a block receipt IBR RPC
> to NN for reporting the block. Even if a DN has received multiple blocks at
> the same time, it still sends multiple RPCs. It does not scale well since NN
> has to process a huge number of RPCs when many DNs receiving many blocks at
> the same time.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
