Uploaded image for project: 'Hadoop HDFS'
  1. Hadoop HDFS
  2. HDFS-918

Use single Selector and small thread pool to replace many instances of BlockSender for reads

    XMLWordPrintableJSON

Details

    • Improvement
    • Status: Open
    • Major
    • Resolution: Unresolved
    • None
    • None
    • datanode, performance
    • None

    Description

      Currently, on read requests, the DataXCeiver server allocates a new thread per request, which must allocate its own buffers and leads to higher-than-optimal CPU and memory usage by the sending threads. If we had a single selector and a small threadpool to multiplex request packets, we could theoretically achieve higher performance while taking up fewer resources and leaving more CPU on datanodes available for mapred, hbase or whatever. This can be done without changing any wire protocols.

      Attachments

        1. hdfs-918-pool.patch
          39 kB
          Jay Booth
        2. hbase-hdfs-benchmarks.ods
          23 kB
          Jay Booth
        3. hdfs-918-branch20-append.patch
          90 kB
          Jay Booth
        4. hdfs-918-TRUNK.patch
          81 kB
          Jay Booth
        5. hdfs-918-branch20.2.patch
          91 kB
          Jay Booth
        6. hdfs-918-20100309.patch
          76 kB
          Jay Booth
        7. hdfs-918-20100228.patch
          73 kB
          Jay Booth
        8. hdfs-918-20100211.patch
          82 kB
          Jay Booth
        9. hdfs-918-20100203.patch
          83 kB
          Jay Booth
        10. hdfs-918-20100201.patch
          81 kB
          Jay Booth
        11. hdfs-multiplex.patch
          83 kB
          Jay Booth

        Issue Links

          depends upon

          Improvement - An improvement or enhancement to an existing feature or task. HDFS-881 Refactor DataNode Packet header into DataTransferProtocol

          • Major - Major loss of function.
          • Closed
          is related to

          Bug - A problem which impairs or prevents the functions of the product. HDFS-2243 DataXceiver per accept seems to be a bottleneck in HBase/YCSB test

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Resolved

          Improvement - An improvement or enhancement to an existing feature or task. HDFS-1599 Umbrella Jira for Improving HBASE support in HDFS

          • Major - Major loss of function.
          • Open
          relates to

          Improvement - An improvement or enhancement to an existing feature or task. HDFS-1034 Enhance datanode to read data and checksum file in parallel

          • Major - Major loss of function.
          • Open

          Improvement - An improvement or enhancement to an existing feature or task. HDFS-1323 Pool/share file channels for HDFS read

          • Major - Major loss of function.
          • Open

          Activity

            People

              jbooth Jay Booth
              jbooth Jay Booth
              Votes:
              2 Vote for this issue
              Watchers:
              39 Start watching this issue

              Dates

                Created:
                Updated: