spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From chenghao-intel <...@git.apache.org>
Subject [GitHub] spark pull request: [SPARK-8813][SQL] Combine files when there're ...
Date Fri, 28 Aug 2015 02:38:40 GMT
Github user chenghao-intel commented on a diff in the pull request:

    https://github.com/apache/spark/pull/8125#discussion_r38167285
  
    --- Diff: sql/core/src/main/scala/org/apache/spark/sql/sources/CombineSmallFile.scala
---
    @@ -0,0 +1,43 @@
    +/*
    + * 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.spark.sql.sources
    +
    +import org.apache.hadoop.fs.{FileStatus, FileSystem, Path}
    +import org.apache.spark.rdd.RDD
    +import org.apache.spark.sql.SQLContext
    +
    +object CombineSmallFile {
    +  def combineWithFiles[T](rdd: RDD[T], sqlContext: SQLContext, inputFiles: Array[FileStatus])
    +      : RDD[T] = {
    +    if (sqlContext.conf.combineSmallFile) {
    +      val totalLen = inputFiles.map { file =>
    +        if (file.isDir) 0L else file.getLen
    +      }.sum
    +      val numPartitions = (totalLen / sqlContext.conf.splitSize + 1).toInt
    +      rdd.coalesce(numPartitions)
    --- End diff --
    
    I think this is a very hack way to solve this problem. As we can not tell how the the
data source to be split, even for Hadoop, the split size just a hint, use that for computing
the partition number probably too risky for a generic data process framework.
    
    And the `RDD.coalesce` actually will combine the splits in a arbitrary way, it's probably
causes the data skew, as we most likely combine the large partitions into a a single task.
    
    IMO, I'd like to deep investigate how Hive to combine the small partitions, by using the
`CombineHiveInputFormat` or `HiveInputFormat`, which seems has a strategy to select the partitions
according to both input format, and also keep the balance.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message