org.apache.hadoop.hbase.io.hfile.bucket.SharedMemoryMmapIOEngine Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of hbase-server Show documentation
Show all versions of hbase-server Show documentation
Server functionality for HBase
/**
* 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.hadoop.hbase.io.hfile.bucket;
import java.io.IOException;
import org.apache.hadoop.hbase.io.hfile.Cacheable;
import org.apache.hadoop.hbase.io.hfile.Cacheable.MemoryType;
import org.apache.hadoop.hbase.io.hfile.CacheableDeserializer;
import org.apache.hadoop.hbase.nio.ByteBuff;
import org.apache.yetus.audience.InterfaceAudience;
/**
* IO engine that stores data in pmem devices such as DCPMM. This engine also mmaps the file from
* the given path. But note that this path has to be a path on the pmem device so that when mmapped
* the file's address is mapped to the Pmem's address space and not in the DRAM. Since this address
* space is exclusive for the Pmem device there is no swapping out of the mmapped contents that
* generally happens when DRAM's free space is not enough to hold the specified file's mmapped
* contents. This gives us the option of using the {@code MemoryType#SHARED} type when serving the
* data from this pmem address space. We need not copy the blocks to the onheap space as we need to
* do for the case of {@code ExclusiveMemoryMmapIOEngine}.
*/
@InterfaceAudience.Private
public class SharedMemoryMmapIOEngine extends FileMmapIOEngine {
// TODO this will support only one path over Pmem. To make use of multiple Pmem devices mounted,
// we need to support multiple paths like files IOEngine. Support later.
public SharedMemoryMmapIOEngine(String filePath, long capacity) throws IOException {
super(filePath, capacity);
}
@Override
public boolean usesSharedMemory() {
return true;
}
@Override
public Cacheable read(long offset, int length, CacheableDeserializer deserializer)
throws IOException {
ByteBuff dstBuffer = bufferArray.asSubByteBuff(offset, length);
// Here the buffer that is created directly refers to the buffer in the actual buckets.
// When any cell is referring to the blocks created out of these buckets then it means that
// those cells are referring to a shared memory area which if evicted by the BucketCache would
// lead to corruption of results. Hence we set the type of the buffer as SHARED_MEMORY
// so that the readers using this block are aware of this fact and do the necessary action
// to prevent eviction till the results are either consumed or copied
return deserializer.deserialize(dstBuffer, true, MemoryType.SHARED);
}
}