Uploaded image for project: 'Hadoop HDFS'
  1. Hadoop HDFS
  2. HDFS-14102

Performance improvement in BlockPlacementPolicyDefault

    XMLWordPrintableJSON

    Details

    • Type: Improvement
    • Status: Resolved
    • Priority: Minor
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 3.3.0
    • Component/s: None
    • Labels:
      None

      Description

       

      BlockPlacementPolicyDefault.java
          // 1. Check that all locations are different.
          // 2. Count locations on different racks.
          Set<String> racks = new TreeSet<>();
          for (DatanodeInfo dn : locs)
            racks.add(dn.getNetworkLocation());
      ...
          racks.size()

       
      Here, the code is counting the number of distinct Network Locations. However, it is using a TreeSet which has overhead to maintain item order and uses a linked structure internally. This overhead is unneeded since all that is required here is a count.

      A NavigableSet implementation based on a TreeMap. The elements are ordered using their natural ordering, or by a Comparator provided at set creation time, depending on which constructor is used.
      This implementation provides guaranteed log time cost for the basic operations (add, remove and contains).

      https://docs.oracle.com/javase/7/docs/api/java/util/TreeSet.html

       
      Use Java streams for readability and because it uses a HashSet under the covers to perform the distinct action. HashSet uses an array internally and has constant time performance for the add method.

      https://github.com/apache/hadoop/blob/27978bcb66a9130cbf26d37ec454c0b7fcdc2530/hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockPlacementPolicyDefault.java#L1042

        Attachments

        1. HDFS-14102.3.patch
          2 kB
          David Mollitor
        2. HDFS-14102.2.patch
          2 kB
          David Mollitor
        3. HDFS-14102.1.patch
          2 kB
          David Mollitor

          Issue Links

            Activity

              People

              • Assignee:
                belugabehr David Mollitor
                Reporter:
                belugabehr David Mollitor
              • Votes:
                0 Vote for this issue
                Watchers:
                4 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved: