yunfengzhou-hub commented on code in PR #97: URL: https://github.com/apache/flink-ml/pull/97#discussion_r874553395
########## flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/FsSegmentWriter.java: ########## @@ -0,0 +1,96 @@ +/* + * 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.flink.iteration.datacache.nonkeyed; + +import org.apache.flink.annotation.Internal; +import org.apache.flink.api.common.typeutils.TypeSerializer; +import org.apache.flink.core.fs.FSDataOutputStream; +import org.apache.flink.core.fs.FileSystem; +import org.apache.flink.core.fs.Path; +import org.apache.flink.core.memory.DataOutputView; +import org.apache.flink.core.memory.DataOutputViewStreamWrapper; + +import java.io.ByteArrayOutputStream; +import java.io.IOException; +import java.io.ObjectOutputStream; +import java.util.Optional; + +/** A class that writes cache data to file system. */ +@Internal +public class FsSegmentWriter<T> implements SegmentWriter<T> { + private final FileSystem fileSystem; + + // TODO: adjust the file size limit automatically according to the provided file system. + private static final int CACHE_FILE_SIZE_LIMIT = 100 * 1024 * 1024; // 100MB + + private final TypeSerializer<T> serializer; + + private final Path path; + + private final FSDataOutputStream outputStream; + + private final ByteArrayOutputStream byteArrayOutputStream; + + private final ObjectOutputStream objectOutputStream; + + private final DataOutputView outputView; + + private int count; + + public FsSegmentWriter(TypeSerializer<T> serializer, Path path) throws IOException { + this.serializer = serializer; + this.path = path; + this.fileSystem = path.getFileSystem(); + this.outputStream = fileSystem.create(path, FileSystem.WriteMode.NO_OVERWRITE); + this.byteArrayOutputStream = new ByteArrayOutputStream(); + this.objectOutputStream = new ObjectOutputStream(outputStream); Review Comment: My experiments show that if we do `outputView = new DataOutputViewStreamWrapper(outputStream)`, the performance of disk IO would decrease obviously. One possible cause is that in `DataOutputViewStreamWrapper` or `FsDataOutputStream`'s subclass, the implementation flushes out bytes every time `write()` is invoked, as I can observe from the update frequency of `outputStream.getPos()`, but I have not dived further into this issue. Shall we leave this implementation as it is for now, and see if @zhipeng93 would offer a more simplified implementation with better performance? I'll leave a TODO here for now. -- 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. To unsubscribe, e-mail: issues-unsubscr...@flink.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
