org.apache.giraph.comm

Class ServerData<I extends org.apache.hadoop.io.WritableComparable,V extends org.apache.hadoop.io.Writable,E extends org.apache.hadoop.io.Writable>

java.lang.Object

org.apache.giraph.comm.ServerData<I,V,E>

Type Parameters:

I - Vertex id

V - Vertex data

E - Edge data

public class ServerData<I extends org.apache.hadoop.io.WritableComparable,V extends org.apache.hadoop.io.Writable,E extends org.apache.hadoop.io.Writable>
extends Object

Anything that the server stores

Constructor Summary

Constructors

Constructor and Description
ServerData(CentralizedServiceWorker<I,V,E> service, WorkerServer workerServer, ImmutableClassesGiraphConfiguration<I,V,E> conf, org.apache.hadoop.mapreduce.Mapper.Context context) Constructor.

Method Summary

Modifier and Type	Method and Description
void	addIncomingWorkerToWorkerMessage(org.apache.hadoop.io.Writable message) Add incoming message to this worker for next superstep.
AllAggregatorServerData	getAllAggregatorData() Get holder for aggregators from previous superstep
List<org.apache.hadoop.io.Writable>	getAndClearCurrentWorkerToWorkerMessages() Get and clear worker to worker messages for this superstep.
<M extends org.apache.hadoop.io.Writable> MessageStore<I,M>	getCurrentMessageStore() Get message store for current messages (messages which we received in previous super step and which will be consumed in current super step)
List<org.apache.hadoop.io.Writable>	getCurrentWorkerToWorkerMessages() Get worker to worker messages received in previous superstep.
EdgeStore<I,V,E>	getEdgeStore() Return the edge store for this worker.
<M extends org.apache.hadoop.io.Writable> MessageStore<I,M>	getIncomingMessageStore() Get message store for incoming messages (messages which will be consumed in the next super step)
OutOfCoreEngine	getOocEngine() Return the out-of-core engine for this worker.
OwnerAggregatorServerData	getOwnerAggregatorData() Get holder for aggregators which current worker owns
ConcurrentMap<Integer,ConcurrentMap<I,VertexMutations<I,V,E>>>	getPartitionMutations() Get the vertex mutations (synchronize on the values)
PartitionStore<I,V,E>	getPartitionStore() Return the partition store for this worker.
CentralizedServiceWorker<I,V,E>	getServiceWorker() Get the reference of the service worker.
void	prepareResolveMutations() Prepare resolving mutation.
void	prepareSuperstep() Prepare for next superstep
void	resetMessageStores() Re-initialize message stores.
void	resolvePartitionMutation(Partition<I,V,E> partition) Resolve mutations specific for a partition.
void	waitForComplete() In case of async message store we have to wait for all messages to be processed before going into next superstep.

Methods inherited from class java.lang.Object

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

Constructor Detail

ServerData

public ServerData(CentralizedServiceWorker<I,V,E> service,
                  WorkerServer workerServer,
                  ImmutableClassesGiraphConfiguration<I,V,E> conf,
                  org.apache.hadoop.mapreduce.Mapper.Context context)

Constructor.

Parameters:

service - Service worker

workerServer - Worker server

conf - Configuration

context - Mapper context

Method Detail

getOocEngine

public OutOfCoreEngine getOocEngine()

Return the out-of-core engine for this worker.

Returns:

The out-of-core engine

getEdgeStore

public EdgeStore<I,V,E> getEdgeStore()

Return the edge store for this worker.

Returns:

The edge store

getPartitionStore

public PartitionStore<I,V,E> getPartitionStore()

Return the partition store for this worker.

Returns:

The partition store

getIncomingMessageStore

public <M extends org.apache.hadoop.io.Writable> MessageStore<I,M> getIncomingMessageStore()

Get message store for incoming messages (messages which will be consumed in the next super step)

Type Parameters:

M - Message data

Returns:

Incoming message store

getCurrentMessageStore

public <M extends org.apache.hadoop.io.Writable> MessageStore<I,M> getCurrentMessageStore()

Get message store for current messages (messages which we received in previous super step and which will be consumed in current super step)

Type Parameters:

M - Message data

Returns:

Current message store

resetMessageStores

public void resetMessageStores()

Re-initialize message stores. Discards old values if any.

prepareSuperstep

public void prepareSuperstep()

Prepare for next superstep

getPartitionMutations

public ConcurrentMap<Integer,ConcurrentMap<I,VertexMutations<I,V,E>>> getPartitionMutations()

Get the vertex mutations (synchronize on the values)

Returns:

Vertex mutations

getOwnerAggregatorData

public OwnerAggregatorServerData getOwnerAggregatorData()

Get holder for aggregators which current worker owns

Returns:

Holder for aggregators which current worker owns

getAllAggregatorData

public AllAggregatorServerData getAllAggregatorData()

Get holder for aggregators from previous superstep

Returns:

Holder for aggregators from previous superstep

getServiceWorker

public CentralizedServiceWorker<I,V,E> getServiceWorker()

Get the reference of the service worker.

Returns:

CentralizedServiceWorker

getAndClearCurrentWorkerToWorkerMessages

public List<org.apache.hadoop.io.Writable> getAndClearCurrentWorkerToWorkerMessages()

Get and clear worker to worker messages for this superstep. Can be called only once per superstep.

Returns:

List of messages for this worker

addIncomingWorkerToWorkerMessage

public void addIncomingWorkerToWorkerMessage(org.apache.hadoop.io.Writable message)

Add incoming message to this worker for next superstep. Thread-safe.

Parameters:

message - Message received

getCurrentWorkerToWorkerMessages

public List<org.apache.hadoop.io.Writable> getCurrentWorkerToWorkerMessages()

Get worker to worker messages received in previous superstep.

Returns:

list of current worker to worker messages.

prepareResolveMutations

public void prepareResolveMutations()

Prepare resolving mutation.

resolvePartitionMutation

public void resolvePartitionMutation(Partition<I,V,E> partition)

Resolve mutations specific for a partition. This method is called once per partition, before the computation for that partition starts.

Parameters:

partition - The partition to resolve mutations for

waitForComplete

public void waitForComplete()

In case of async message store we have to wait for all messages to be processed before going into next superstep.