org.apache.flink.table.runtime.hashtable

Class BaseHybridHashTable

java.lang.Object

org.apache.flink.table.runtime.hashtable.BaseHybridHashTable

All Implemented Interfaces:

org.apache.flink.core.memory.MemorySegmentSource, MemorySegmentPool

Direct Known Subclasses:

BinaryHashTable, LongHybridHashTable

public abstract class BaseHybridHashTable
extends Object
implements MemorySegmentPool

Base table for LongHybridHashTable and BinaryHashTable.

Field Summary

Fields

Modifier and Type	Field and Description
protected long	buildRowCount
protected int	buildSpillRetBufferNumbers The number of buffers in the build spill return buffer queue that are actually not write behind buffers, but regular buffers that only have not yet returned.
protected LinkedBlockingQueue<org.apache.flink.core.memory.MemorySegment>	buildSpillReturnBuffers The queue of buffers that can be used for write-behind.
protected AtomicBoolean	closed Flag indicating that the closing logic has been invoked.
protected int	compressionBlockSize
protected org.apache.flink.runtime.io.compression.BlockCompressionFactory	compressionCodecFactory
protected boolean	compressionEnabled
protected org.apache.flink.runtime.io.disk.iomanager.FileIOChannel.Enumerator	currentEnumerator The channel enumerator that is used while processing the current partition to create channels for the spill partitions it requires.
protected int	currentRecursionDepth The recursion depth of the partition that is currently processed.
protected org.apache.flink.runtime.io.disk.iomanager.HeaderlessChannelReaderInputView	currentSpilledBuildSide The reader for the spilled-file of the build partition that is currently read.
protected org.apache.flink.runtime.io.disk.iomanager.AbstractChannelReaderInputView	currentSpilledProbeSide The reader for the spilled-file of the probe partition that is currently read.
protected int	initPartitionFanOut
protected LazyMemorySegmentPool	internalPool
protected org.apache.flink.runtime.io.disk.iomanager.IOManager	ioManager The I/O manager used to instantiate writers for the spilled partitions.
protected static org.slf4j.Logger	LOG
protected static int	MAX_NUM_PARTITIONS The maximum number of partitions, which defines the spilling granularity.
protected static int	MAX_RECURSION_DEPTH The maximum number of recursive partitionings that the join does before giving up.
protected long	numSpillFiles
protected int	segmentSize The size of the segments used by the hash join buckets.
int	segmentSizeBits
int	segmentSizeMask
protected long	spillInBytes
protected int	totalNumBuffers The total reserved number of memory segments available to the hash join.
boolean	tryDistinctBuildRow Try to make the buildSide rows distinct.

Constructor Summary

Constructors

Constructor and Description
BaseHybridHashTable(Object owner, boolean compressionEnabled, int compressionBlockSize, org.apache.flink.runtime.memory.MemoryManager memManager, long reservedMemorySize, org.apache.flink.runtime.io.disk.iomanager.IOManager ioManager, int avgRecordLen, long buildRowCount, boolean tryDistinctBuildRow)

Method Summary

Modifier and Type	Method and Description
protected abstract void	clearPartitions()
void	close() Closes the hash table.
protected org.apache.flink.runtime.io.disk.iomanager.HeaderlessChannelReaderInputView	createInputView(org.apache.flink.runtime.io.disk.iomanager.FileIOChannel.ID id, int blockCount, int lastSegmentLimit)
void	ensureNumBuffersReturned(int minRequiredAvailable) This method makes sure that at least a certain number of memory segments is in the list of free segments.
void	free()
void	freeCurrent() Free the memory not used.
int	freePages()
org.apache.flink.core.memory.MemorySegment	getNextBuffer() Gets the next buffer to be used with the hash-table, either for an in-memory partition, or for the table buckets.
org.apache.flink.core.memory.MemorySegment[]	getNextBuffers(int bufferSize) Bulk memory acquisition.
protected org.apache.flink.core.memory.MemorySegment	getNotNullNextBuffer()
long	getNumSpillFiles()
long	getSpillInBytes()
long	getUsedMemoryInBytes()
static int	hash(int hashCode, int level) The level parameter is needed so that we can have different hash functions when we recursively apply the partitioning, so that the working set eventually fits into memory.
int	maxInitBufferOfBucketArea(int partitions) Give up to one-sixth of the memory of the bucket area.
protected int	maxNumPartition() Bucket area need at-least one and data need at-least one.
org.apache.flink.core.memory.MemorySegment	nextSegment() This is the method called by the partitions to request memory to serialize records.
int	pageSize() Get the page size of each page this pool holds.
protected List<org.apache.flink.core.memory.MemorySegment>	readAllBuffers(org.apache.flink.runtime.io.disk.iomanager.FileIOChannel.ID id, int blockCount)
void	releaseMemoryCacheForSMJ() Due to adaptive hash join is introduced, the cached memory segments should be released to MemoryManager before switch to sort merge join.
int	remainBuffers()
void	returnAll(List<org.apache.flink.core.memory.MemorySegment> memory) Return all pages back into this pool.
void	returnPage(org.apache.flink.core.memory.MemorySegment segment)
protected abstract int	spillPartition()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

LOG

protected static final org.slf4j.Logger LOG

MAX_RECURSION_DEPTH

protected static final int MAX_RECURSION_DEPTH

The maximum number of recursive partitionings that the join does before giving up.

See Also:

Constant Field Values

MAX_NUM_PARTITIONS

protected static final int MAX_NUM_PARTITIONS

The maximum number of partitions, which defines the spilling granularity. Each recursion, the data is divided maximally into that many partitions, which are processed in one chuck.

See Also:

Constant Field Values

initPartitionFanOut

protected final int initPartitionFanOut

buildRowCount

protected final long buildRowCount

totalNumBuffers

protected final int totalNumBuffers

The total reserved number of memory segments available to the hash join.

internalPool

protected final LazyMemorySegmentPool internalPool

ioManager

protected final org.apache.flink.runtime.io.disk.iomanager.IOManager ioManager

The I/O manager used to instantiate writers for the spilled partitions.

segmentSize

protected final int segmentSize

The size of the segments used by the hash join buckets. All segments must be of equal size to ease offset computations.

buildSpillReturnBuffers

protected final LinkedBlockingQueue<org.apache.flink.core.memory.MemorySegment> buildSpillReturnBuffers

The queue of buffers that can be used for write-behind. Any buffer that is written asynchronously to disk is returned through this queue. hence

segmentSizeBits

public final int segmentSizeBits

segmentSizeMask

public final int segmentSizeMask

closed

protected AtomicBoolean closed

Flag indicating that the closing logic has been invoked.

tryDistinctBuildRow

public final boolean tryDistinctBuildRow

Try to make the buildSide rows distinct.

currentRecursionDepth

protected int currentRecursionDepth

The recursion depth of the partition that is currently processed. The initial table has a recursion depth of 0. Partitions spilled from a table that is built for a partition with recursion depth n have a recursion depth of n+1.

buildSpillRetBufferNumbers

protected int buildSpillRetBufferNumbers

The number of buffers in the build spill return buffer queue that are actually not write behind buffers, but regular buffers that only have not yet returned. This is part of an optimization that the spilling code needs not wait until the partition is completely spilled before proceeding.

currentSpilledBuildSide

protected org.apache.flink.runtime.io.disk.iomanager.HeaderlessChannelReaderInputView currentSpilledBuildSide

The reader for the spilled-file of the build partition that is currently read.

currentSpilledProbeSide

protected org.apache.flink.runtime.io.disk.iomanager.AbstractChannelReaderInputView currentSpilledProbeSide

The reader for the spilled-file of the probe partition that is currently read.

currentEnumerator

protected org.apache.flink.runtime.io.disk.iomanager.FileIOChannel.Enumerator currentEnumerator

The channel enumerator that is used while processing the current partition to create channels for the spill partitions it requires.

compressionEnabled

protected final boolean compressionEnabled

compressionCodecFactory

protected final org.apache.flink.runtime.io.compression.BlockCompressionFactory compressionCodecFactory

compressionBlockSize

protected final int compressionBlockSize

numSpillFiles

protected transient long numSpillFiles

spillInBytes

protected transient long spillInBytes

Constructor Detail

BaseHybridHashTable

public BaseHybridHashTable(Object owner,
                           boolean compressionEnabled,
                           int compressionBlockSize,
                           org.apache.flink.runtime.memory.MemoryManager memManager,
                           long reservedMemorySize,
                           org.apache.flink.runtime.io.disk.iomanager.IOManager ioManager,
                           int avgRecordLen,
                           long buildRowCount,
                           boolean tryDistinctBuildRow)

Method Detail

maxNumPartition

protected int maxNumPartition()

Bucket area need at-least one and data need at-least one. In the initialization phase, we can use (totalNumBuffers - numWriteBehindBuffers) Segments. However, in the buildTableFromSpilledPartition phase, only (totalNumBuffers - numWriteBehindBuffers - 2) can be used because two Buffers are needed to read the data.

getNextBuffer

public org.apache.flink.core.memory.MemorySegment getNextBuffer()

Gets the next buffer to be used with the hash-table, either for an in-memory partition, or for the table buckets. This method returns null, if no more buffer is available. Spilling a partition may free new buffers then.

Returns:

The next buffer to be used by the hash-table, or null, if no buffer remains.

getNextBuffers

public org.apache.flink.core.memory.MemorySegment[] getNextBuffers(int bufferSize)

Bulk memory acquisition. NOTE: Failure to get memory will throw an exception.

getNotNullNextBuffer

protected org.apache.flink.core.memory.MemorySegment getNotNullNextBuffer()

nextSegment

public org.apache.flink.core.memory.MemorySegment nextSegment()

This is the method called by the partitions to request memory to serialize records. It automatically spills partitions, if memory runs out.

Specified by:

nextSegment in interface org.apache.flink.core.memory.MemorySegmentSource

Returns:

The next available memory segment.

freePages

public int freePages()

Specified by:

freePages in interface MemorySegmentPool

Returns:

Free page number.

pageSize

public int pageSize()

Description copied from interface: MemorySegmentPool

Get the page size of each page this pool holds.

Specified by:

pageSize in interface MemorySegmentPool

Returns:

the page size

returnAll

public void returnAll(List<org.apache.flink.core.memory.MemorySegment> memory)

Description copied from interface: MemorySegmentPool

Return all pages back into this pool.

Specified by:

returnAll in interface MemorySegmentPool

Parameters:

memory - the pages which want to be returned.

spillPartition

protected abstract int spillPartition()
                               throws IOException

Throws:

IOException

ensureNumBuffersReturned

public void ensureNumBuffersReturned(int minRequiredAvailable)

This method makes sure that at least a certain number of memory segments is in the list of free segments. Free memory can be in the list of free segments, or in the return-queue where segments used to write behind are put. The number of segments that are in that return-queue, but are actually reclaimable is tracked. This method makes sure at least a certain number of buffers is reclaimed.

Parameters:

minRequiredAvailable - The minimum number of buffers that needs to be reclaimed.

close

public void close()

Closes the hash table. This effectively releases all internal structures and closes all open files and removes them. The call to this method is valid both as a cleanup after the complete inputs were properly processed, and as an cancellation call, which cleans up all resources that are currently held by the hash join.

clearPartitions

protected abstract void clearPartitions()

free

public void free()

freeCurrent

public void freeCurrent()

Free the memory not used.

releaseMemoryCacheForSMJ

public void releaseMemoryCacheForSMJ()

Due to adaptive hash join is introduced, the cached memory segments should be released to MemoryManager before switch to sort merge join. Otherwise, open sort merge join operator maybe fail because of insufficient memory.

Note: this method should only be invoked for sort merge join.

returnPage

public void returnPage(org.apache.flink.core.memory.MemorySegment segment)

remainBuffers

public int remainBuffers()

getUsedMemoryInBytes

public long getUsedMemoryInBytes()

getNumSpillFiles

public long getNumSpillFiles()

getSpillInBytes

public long getSpillInBytes()

maxInitBufferOfBucketArea

public int maxInitBufferOfBucketArea(int partitions)

Give up to one-sixth of the memory of the bucket area.

readAllBuffers

protected List<org.apache.flink.core.memory.MemorySegment> readAllBuffers(org.apache.flink.runtime.io.disk.iomanager.FileIOChannel.ID id,
                                                                          int blockCount)
                                                                   throws IOException

Throws:

IOException

createInputView

protected org.apache.flink.runtime.io.disk.iomanager.HeaderlessChannelReaderInputView createInputView(org.apache.flink.runtime.io.disk.iomanager.FileIOChannel.ID id,
                                                                                                      int blockCount,
                                                                                                      int lastSegmentLimit)
                                                                                               throws IOException

Throws:

IOException

hash

public static int hash(int hashCode,
                       int level)

The level parameter is needed so that we can have different hash functions when we recursively apply the partitioning, so that the working set eventually fits into memory.