You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@pinot.apache.org by GitBox <gi...@apache.org> on 2020/03/31 05:55:39 UTC
[GitHub] [incubator-pinot] siddharthteotia opened a new pull request #5199:
Remove the construction of second bitmap in text index reader to improve
performance
siddharthteotia opened a new pull request #5199: Remove the construction of second bitmap in text index reader to improve performance
URL: https://github.com/apache/incubator-pinot/pull/5199
This is a follow-up to optimization implemented in PR https://github.com/apache/incubator-pinot/pull/5177.
Since we now have pre-built mapping of luceneDocId to pinotDocId, we can directly build the result bitmap with pinotDocId. This PR removes the construction of second bitmap since earlier we had to do build the result in two phases -- (1) run search query to get luceneDocIDs in a bitmap. Iterate over this bitmap and build a second bitmap with corresponding pinotDocIds.
Now in our Lucene collector callback, we can directly build the final bitmap.
This change along with previous PR provides significant performance improvements.
Ran an increasingQPS test on real data (single segment with 10million docs and a text index). QPS was increased from 1 to 40
`REPORT FOR TARGET QPS: 1.0
Current Target QPS: 1.0, Time Passed: 30084ms, Queries Executed: 30, Average QPS: 0.997207818109294, Average Broker Time: 44.6ms, Average Client Time: 50.1ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 3.0
Current Target QPS: 3.0, Time Passed: 30255ms, Queries Executed: 90, Average QPS: 2.974714923153198, Average Broker Time: 50.3ms, Average Client Time: 53.24444444444445ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 5.0
Current Target QPS: 5.0, Time Passed: 30412ms, Queries Executed: 150, Average QPS: 4.932263580165724, Average Broker Time: 41.44ms, Average Client Time: 44.1ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 7.0
Current Target QPS: 7.0, Time Passed: 30489ms, Queries Executed: 210, Average QPS: 6.887730000983961, Average Broker Time: 41.82857142857143ms, Average Client Time: 44.00476190476191ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 9.0
Current Target QPS: 9.0, Time Passed: 30868ms, Queries Executed: 270, Average QPS: 8.746922379162887, Average Broker Time: 43.385185185185186ms, Average Client Time: 45.27037037037037ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 25.0
Current Target QPS: 25.0, Time Passed: 30233ms, Queries Executed: 694, Average QPS: 22.955049118512882, Average Broker Time: 37.27089337175793ms, Average Client Time: 38.53746397694525ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 27.0
Current Target QPS: 27.0, Time Passed: 30254ms, Queries Executed: 740, Average QPS: 24.459575593309975, Average Broker Time: 39.71351351351351ms, Average Client Time: 40.87837837837838ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 29.0
Current Target QPS: 29.0, Time Passed: 30147ms, Queries Executed: 798, Average QPS: 26.4702955517962, Average Broker Time: 37.06516290726817ms, Average Client Time: 38.22431077694235ms, Queries Queued: 0.
REPORT FOR TARGET QPS: 31.0
Current Target QPS: 31.0, Time Passed: 30160ms, Queries Executed: 843, Average QPS: 27.950928381962864, Average Broker Time: 37.79359430604982ms, Average Client Time: 38.86476868327402ms, Queries Queued: 0.
FINAL REPORT:
Current Target QPS: 39.0, Time Passed: 27344ms, Queries Executed: 947, Average QPS: 34.632826214160325, Average Broker Time: 36.91024287222809ms, Average Client Time: 37.91129883843717ms.
10th percentile: 4.0
25th percentile: 26.0
50th percentile: 37.0
90th percentile: 68.0
95th percentile: 75.0
99th percentile: 129.0
99.9th percentile: 150.0`
The optimizations implemented in this and previous PR are not directly applicable to realtime (we have the exact same performance overhead in realtime too) since we can't have a pre-built mapping there. We mostly need to build a cache on-the-fly as queries are processed on realtime lucene index. A solution is in progress. Will put PR soon
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@pinot.apache.org
For additional commands, e-mail: commits-help@pinot.apache.org
[GitHub] [incubator-pinot] siddharthteotia commented on a change in pull
request #5199: Remove the construction of second bitmap in text index
reader to improve performance
Posted by GitBox <gi...@apache.org>.
siddharthteotia commented on a change in pull request #5199: Remove the construction of second bitmap in text index reader to improve performance
URL: https://github.com/apache/incubator-pinot/pull/5199#discussion_r401161432
##########
File path: pinot-core/src/main/java/org/apache/pinot/core/realtime/impl/invertedindex/RealtimeLuceneDocIdCollector.java
##########
@@ -0,0 +1,64 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+package org.apache.pinot.core.realtime.impl.invertedindex;
+
+import java.io.IOException;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.search.Collector;
+import org.apache.lucene.search.LeafCollector;
+import org.apache.lucene.search.Scorable;
+import org.apache.lucene.search.ScoreMode;
+import org.roaringbitmap.buffer.MutableRoaringBitmap;
+
+/**
+ * DocID collector for Lucene search query. We have optimized
Review comment:
> (OCD) Can we always use `docId` for naming instead of mixing use of `docId` and `docID`
Done.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@pinot.apache.org
For additional commands, e-mail: commits-help@pinot.apache.org
[GitHub] [incubator-pinot] siddharthteotia merged pull request #5199: Remove
the construction of second bitmap in text index reader to improve performance
Posted by GitBox <gi...@apache.org>.
siddharthteotia merged pull request #5199: Remove the construction of second bitmap in text index reader to improve performance
URL: https://github.com/apache/incubator-pinot/pull/5199
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@pinot.apache.org
For additional commands, e-mail: commits-help@pinot.apache.org
[GitHub] [incubator-pinot] Jackie-Jiang commented on a change in pull
request #5199: Remove the construction of second bitmap in text index
reader to improve performance
Posted by GitBox <gi...@apache.org>.
Jackie-Jiang commented on a change in pull request #5199: Remove the construction of second bitmap in text index reader to improve performance
URL: https://github.com/apache/incubator-pinot/pull/5199#discussion_r401125829
##########
File path: pinot-core/src/main/java/org/apache/pinot/core/realtime/impl/invertedindex/RealtimeLuceneDocIdCollector.java
##########
@@ -0,0 +1,64 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+package org.apache.pinot.core.realtime.impl.invertedindex;
+
+import java.io.IOException;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.search.Collector;
+import org.apache.lucene.search.LeafCollector;
+import org.apache.lucene.search.Scorable;
+import org.apache.lucene.search.ScoreMode;
+import org.roaringbitmap.buffer.MutableRoaringBitmap;
+
+/**
+ * DocID collector for Lucene search query. We have optimized
Review comment:
(OCD) Can we always use `docId` for naming instead of mixing use of `docId` and `docID`
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@pinot.apache.org
For additional commands, e-mail: commits-help@pinot.apache.org
[GitHub] [incubator-pinot] siddharthteotia commented on a change in pull
request #5199: Remove the construction of second bitmap in text index
reader to improve performance
Posted by GitBox <gi...@apache.org>.
siddharthteotia commented on a change in pull request #5199: Remove the construction of second bitmap in text index reader to improve performance
URL: https://github.com/apache/incubator-pinot/pull/5199#discussion_r401161215
##########
File path: pinot-core/src/main/java/org/apache/pinot/core/realtime/impl/invertedindex/RealtimeLuceneDocIdCollector.java
##########
@@ -0,0 +1,64 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+package org.apache.pinot.core.realtime.impl.invertedindex;
+
+import java.io.IOException;
+import org.apache.lucene.index.LeafReaderContext;
+import org.apache.lucene.search.Collector;
+import org.apache.lucene.search.LeafCollector;
+import org.apache.lucene.search.Scorable;
+import org.apache.lucene.search.ScoreMode;
+import org.roaringbitmap.buffer.MutableRoaringBitmap;
+
+/**
+ * DocID collector for Lucene search query. We have optimized
Review comment:
> I would recommend implementing the real-time docId translator instead of having a separate real-time collector. You can also do that as the next step.
> LGTM otherwise.
Yes, that is the next step (implementing this optimization for realtime)
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@pinot.apache.org
For additional commands, e-mail: commits-help@pinot.apache.org