lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joel Bernstein (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (SOLR-9938) Improve the performance of CloudSolrStream
Date Fri, 06 Jan 2017 16:54:58 GMT

     [ https://issues.apache.org/jira/browse/SOLR-9938?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Joel Bernstein updated SOLR-9938:
---------------------------------
    Description: 
Now that we have started to improve the efficiency of Streaming, I think it makes sense to
work on CloudSolrStream, which is used as a bulk stream source.

The first thing to tackle is how the merge sort of the SolrStreams from each shard is done.

Currently the sorting is done by a TreeSet, which is not the most efficient approach. For
one thing each *put* and *poll* on the TreeSet creates a new map Entry. When streaming millions
of documents this adds up. Also the TreeSet is backed by a TreeMap that maintains a fully
order set of tuples. We just need to know the highest Tuple.

I think we can increase throughput significantly by using a custom priority queue for sorting
rather then the TreeSet.  




  was:
Now that we have started to improve the efficiency of Streaming, I think it makes sense to
work on CloudSolrStream, which is used as a bulk stream source.

The first thing to tackle is how the merge sort of the SolrStream from each shard is done.

Currently the sorting is done by a TreeSet, which is not the most efficient approach. For
one thing each *put* and *poll* on the TreeSet creates a new map Entry. When streaming millions
of documents this adds up. Also the TreeSet is backed by a TreeMap that maintains a fully
order set of tuples. We just need to know the highest Tuple.

I think we can increase throughput significantly by using a custom priority queue for sorting
rather then the TreeSet.  





> Improve the performance of CloudSolrStream
> ------------------------------------------
>
>                 Key: SOLR-9938
>                 URL: https://issues.apache.org/jira/browse/SOLR-9938
>             Project: Solr
>          Issue Type: Improvement
>      Security Level: Public(Default Security Level. Issues are Public) 
>            Reporter: Joel Bernstein
>
> Now that we have started to improve the efficiency of Streaming, I think it makes sense
to work on CloudSolrStream, which is used as a bulk stream source.
> The first thing to tackle is how the merge sort of the SolrStreams from each shard is
done.
> Currently the sorting is done by a TreeSet, which is not the most efficient approach.
For one thing each *put* and *poll* on the TreeSet creates a new map Entry. When streaming
millions of documents this adds up. Also the TreeSet is backed by a TreeMap that maintains
a fully order set of tuples. We just need to know the highest Tuple.
> I think we can increase throughput significantly by using a custom priority queue for
sorting rather then the TreeSet.  



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message