[ 
https://issues.apache.org/jira/browse/CASSANDRA-1958?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12979934#action_12979934
 ] 

lichenglin commented on CASSANDRA-1958:
---------------------------------------

Three nodes and N=3,R=W=2.Get CPU info when we get 10,000 from one key  through 
JPROFILER.

99.4% - 968 s - 103 inv. java.util.concurrent.ThreadPoolExecutor$Worker.run
  99.4% - 968 s - 713 inv. org.apache.cassandra.net.MessageDeliveryTask.run
   99.2% - 966 s - 70 inv. org.apache.cassandra.db.ReadVerbHandler.doVerb
    99.2% - 966 s - 70 inv. org.apache.cassandra.db.SliceFromReadCommand.getRow
     99.2% - 966 s - 70 inv. org.apache.cassandra.db.Table.getRow
      99.2% - 966 s - 70 inv. 
org.apache.cassandra.db.ColumnFamilyStore.getColumnFamily
       99.2% - 966 s - 70 inv. 
org.apache.cassandra.db.ColumnFamilyStore.getColumnFamily
        97.8% - 952 s - 70 inv. 
org.apache.cassandra.db.ColumnFamilyStore.getTopLevelColumns
         93.6% - 912 s - 70 inv. 
org.apache.cassandra.db.filter.QueryFilter.collectCollatedColumns
          93.6% - 912 s - 70 inv. 
org.apache.cassandra.db.filter.SliceQueryFilter.collectReducedColumns
           -75.8% - 739 s - 647,235 inv. java.util.Iterator.hasNext
              -75.6% - 736 s - 647,235 inv. 
com.google.common.collect.AbstractIterator.hasNext
                -75.2% - 733 s - 647,235 inv. 
com.google.common.collect.AbstractIterator.tryToComputeNext
                -0.1% - 691 ms - 647,235 inv. 
com.google.common.base.Preconditions.checkState
           -14.5% - 141 s - 647,165 inv. 
org.apache.cassandra.db.ColumnFamily.addColumn
 
Iterator.hasNext cost 75.8%.

> query slow in million columns of one key
> ----------------------------------------
>
>                 Key: CASSANDRA-1958
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1958
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: 0.6.5
>         Environment: three nodes with SUSE and 2G JVM.
>            Reporter: lichenglin
>
> We get all records in 200 ms with 10,000 columns of one key, but 4 seconds  
> with 100,000 columns, and we fear for 1,000,000.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to