[HBASE-2208] TableServers # processBatchOfRows - converts from List to [ ] - Expensive copy

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.90.0
• Hadoop Flags: Reviewed
• Assignee: Michael Stack
• Reporter: Karthik K
• Created: 10/Feb/10 01:56
• Updated: 20/Nov/15 13:01
• Resolved: 11/Jun/10 18:10

Description

With autoFlush to false and a large write buffer on HTable, when we write bulk puts - TableServer # processBatchOfRows , convert the input (List) to an [ ] , before sending down the wire.

With a write buffer as large as 20 MB , that becomes an expensive copy when we do - list.toArray(new T[ ] ).

May be - should we change the wire protocol to support List as well , and then revisit this to prevent the bulk copy ?

Batch b = new Batch(this) {
        @Override
        int doCall(final List<Row> currentList, final byte [] row,
          final byte [] tableName)
        throws IOException, RuntimeException {
          *final Put [] puts = currentList.toArray(PUT_ARRAY_TYPE);*
          return getRegionServerWithRetries(new ServerCallable<Integer>(this.c,
              tableName, row) {
            public Integer call() throws IOException {
              return server.put(location.getRegionInfo().getRegionName(), puts);
            }
          });
        }

Issue Links

• is related to: HBASE-1845 MultiGet, MultiDelete, and MultiPut - batched to the appropriate region servers