com.bigdata.io.writecache

Class WriteCache.FileChannelScatteredWriteCache

java.lang.Object

com.bigdata.io.writecache.WriteCache

com.bigdata.io.writecache.WriteCache.FileChannelScatteredWriteCache

All Implemented Interfaces:

IWriteCache

Enclosing class:

WriteCache

public static class WriteCache.FileChannelScatteredWriteCache
extends WriteCache

The scattered write cache is used by the RWStore since the writes can be made to any part of the file assigned for data allocation.

The writeonChannel must therefore utilize the WriteCache.RecordMetadata to write each update separately.

To support HA, we prefix each write with the file position and buffer length in the cache. This enables the cache buffer to be sent as a single stream and the RecordMap rebuilt downstream.

Nested Class Summary

Nested classes/interfaces inherited from class com.bigdata.io.writecache.WriteCache

WriteCache.FileChannelScatteredWriteCache, WriteCache.FileChannelWriteCache, WriteCache.HAPackage, WriteCache.ReadCache, WriteCache.RecordMetadata

Field Summary

Fields inherited from class com.bigdata.io.writecache.WriteCache

counters, log, recordMap

Constructor Summary

Constructors

Constructor and Description
WriteCache.FileChannelScatteredWriteCache(IBufferAccess buf, boolean useChecksum, boolean isHighlyAvailable, boolean bufferHasData, IReopenChannel<FileChannel> opener, long fileExtent, BufferedWrite bufferedWrite)

Method Summary

Methods

Modifier and Type	Method and Description
protected void	addAddress(int latchedAddr, int size) A record add has been decoded.
protected void	removeAddress(int latchedAddr) A record delete has been decoded.
void	resetRecordMapFromBuffer(ByteBuffer buf, Map<Long,WriteCache.RecordMetadata> recordMap) Hook to rebuild WriteCache.RecordMetadata after buffer has been transferred.
protected boolean	writeOnChannel(ByteBuffer data, long firstOffsetIgnored, Map<Long,WriteCache.RecordMetadata> recordMap, long nanos) Called by WriteCacheService to process a direct write for large blocks and also to flush data from dirty caches.

Methods inherited from class com.bigdata.io.writecache.WriteCache

allocate, bytesWritten, capacity, close, closeForWrites, contains, decrementReferenceCount, flush, flush, getCompressorKey, getCounters, getFileExtent, getLastOffset, getReferenceCount, incrementReferenceCount, isClosedForWrites, read, registerWriteStatus, reset, resetRecordMapFromBuffer, setFileExtent, setFirstOffset, setRecordMap, toString, write

Methods inherited from class java.lang.Object

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

Constructor Detail

WriteCache.FileChannelScatteredWriteCache

public WriteCache.FileChannelScatteredWriteCache(IBufferAccess buf,
                                         boolean useChecksum,
                                         boolean isHighlyAvailable,
                                         boolean bufferHasData,
                                         IReopenChannel<FileChannel> opener,
                                         long fileExtent,
                                         BufferedWrite bufferedWrite)
                                          throws InterruptedException

Parameters:

baseOffset - An offset

buf -

opener -

Throws:

InterruptedException

Method Detail

writeOnChannel

protected boolean writeOnChannel(ByteBuffer data,
                     long firstOffsetIgnored,
                     Map<Long,WriteCache.RecordMetadata> recordMap,
                     long nanos)
                          throws InterruptedException,
                                 IOException

Called by WriteCacheService to process a direct write for large blocks and also to flush data from dirty caches. TODO The [nanos] parameter is ignored.

Specified by:

writeOnChannel in class WriteCache

Parameters:

data - The data to be written. Only the dirty bytes are visible in this view. The implementation should write all bytes from the current position to the limit.

firstOffsetIgnored - The offset of the first record in the recordMap into the file (may be relative to a base offset within the file). This is provided as an optimization for the WORM which writes its records contiguously on the backing store.

recordMap - The mapping of record offsets onto metadata about those records.

nanos - The timeout for the operation in nanoseconds.

Returns:

true if the operation was completed successfully within the time alloted.

Throws:

InterruptedException - if the thread was interrupted.

IOException - if there was an IO problem.

resetRecordMapFromBuffer

public void resetRecordMapFromBuffer(ByteBuffer buf,
                            Map<Long,WriteCache.RecordMetadata> recordMap)

Hook to rebuild WriteCache.RecordMetadata after buffer has been transferred. For the WriteCache.FileChannelScatteredWriteCache this means hopping trough the buffer marking offsets and data size into the WriteCache.RecordMetadata map, and ignoring any zero address entries that indicate a "freed" allocation.

Update: This has now been changed to avoid problems with incremental checksums by indicating removal by appending a new prefix where the data length is zero.

Overrides:

resetRecordMapFromBuffer in class WriteCache

Throws:

InterruptedException

addAddress

protected void addAddress(int latchedAddr,
              int size)

A record add has been decoded.

Parameters:

latchedAddr - The latched address.

size - The size of the allocation in bytes.

removeAddress

protected void removeAddress(int latchedAddr)

A record delete has been decoded.

Parameters:

latchedAddr - The latched address.