org.apache.ignite.spi.collision.priorityqueue

Class PriorityQueueCollisionSpi

java.lang.Object

org.apache.ignite.spi.IgniteSpiAdapter

org.apache.ignite.spi.collision.priorityqueue.PriorityQueueCollisionSpi

All Implemented Interfaces:

CollisionSpi, IgniteSpi

@IgniteSpiMultipleInstancesSupport(value=true)
 @IgniteSpiConsistencyChecked(optional=true)
public class PriorityQueueCollisionSpi
extends IgniteSpiAdapter
implements CollisionSpi

This class provides implementation for Collision SPI based on priority queue.

Description

Jobs are first ordered by their priority, if one is specified, and only first getParallelJobsNumber() jobs is allowed to execute in parallel. Other jobs will be queued up.

Configuration

Mandatory

This SPI has no mandatory configuration parameters.

Optional

This SPI has the following optional configuration parameters:

• Number of jobs that can be executed in parallel (see setParallelJobsNumber(int)). This number should usually be set to no greater than number of threads in the execution thread pool.
• Priority attribute session key (see getPriorityAttributeKey()). Prior to returning from ComputeTask.map(List, Object) method, task implementation should set a value into the task session keyed by this attribute key. See ComputeTaskSession for more information about task session.
• Priority attribute job context key (see getJobPriorityAttributeKey()). It is used for specifying job priority. See ComputeJobContext for more information about job context.
• Default priority value (see getDefaultPriority()). It is used when no priority is set.
• Default priority increase value (see getStarvationIncrement()). It is used for increasing priority when job gets bumped down. This future is used for preventing starvation waiting jobs execution.
• Default increasing priority flag value (see isStarvationPreventionEnabled()). It is used for enabling increasing priority when job gets bumped down. This future is used for preventing starvation waiting jobs execution.

Below is a Java example of configuration for priority collision SPI:

 PriorityQueueCollisionSpi colSpi = new PriorityQueueCollisionSpi();

 // Execute all jobs sequentially by setting parallel job number to 1.
 colSpi.setParallelJobsNumber(1);

 IgniteConfiguration cfg = new IgniteConfiguration();

 // Override default collision SPI.
 cfg.setCollisionSpi(colSpi);

 // Starts grid.
 G.start(cfg);
 

Here is Spring XML configuration example:

 <property name="collisionSpi">
     <bean class="org.apache.ignite.spi.collision.priorityqueue.PriorityQueueCollisionSpi">
         <property name="priorityAttributeKey" value="myPriorityAttributeKey"/>
         <property name="parallelJobsNumber" value="10"/>
     </bean>
 </property>
 

Coding Example

Here is an example of a grid tasks that uses priority collision SPI configured in example above. Note that priority collision resolution is absolutely transparent to the user and is simply a matter of proper grid configuration. Also, priority may be defined only for task (it can be defined within the task, not at a job level). All split jobs will be started with priority declared in their owner task.

This example demonstrates how urgent task may be declared with a higher priority value. Priority SPI guarantees (see its configuration in example above, where number of parallel jobs is set to 1) that all jobs from MyGridUrgentTask will most likely be activated first (one by one) and jobs from MyGridUsualTask with lowest priority will wait. Once higher priority jobs complete, lower priority jobs will be scheduled.

 public class MyGridUsualTask extends ComputeTaskSplitAdapter<Object, Object> {
    public static final int SPLIT_COUNT = 20;

    @TaskSessionResource
    private ComputeTaskSession taskSes;

    @Override
    protected Collection<? extends ComputeJob> split(int gridSize, Object arg) throws IgniteCheckedException {
        ...
        // Set low task priority (note that attribute name is used by the SPI
        // and should not be changed).
        taskSes.setAttribute("grid.task.priority", 5);

        Collection<ComputeJob> jobs = new ArrayList<ComputeJob>(SPLIT_COUNT);

        for (int i = 1; i <= SPLIT_COUNT; i++) {
            jobs.add(new ComputeJobAdapter<Integer>(i) {
                ...
            });
        }
        ...
    }
 }
 

and

 public class MyGridUrgentTask extends ComputeTaskSplitAdapter<Object, Object> {
    public static final int SPLIT_COUNT = 5;

    @TaskSessionResource
    private ComputeTaskSession taskSes;

    @Override
    protected Collection<? extends ComputeJob> split(int gridSize, Object arg) throws IgniteCheckedException {
        ...
        // Set high task priority (note that attribute name is used by the SPI
        // and should not be changed).
        taskSes.setAttribute("grid.task.priority", 10);

        Collection<ComputeJob> jobs = new ArrayList<ComputeJob>(SPLIT_COUNT);

        for (int i = 1; i <= SPLIT_COUNT; i++) {
            jobs.add(new ComputeJobAdapter<Integer>(i) {
                ...
            });
        }
        ...
    }
 }
 

For information about Spring framework visit www.springframework.org

Field Summary

Fields

Modifier and Type	Field and Description
static String	DFLT_JOB_PRIORITY_ATTRIBUTE_KEY Default job priority attribute key (value is grid.job.priority).
static int	DFLT_PARALLEL_JOBS_NUM Default number of parallel jobs allowed (set to number of cores times 2).
static boolean	DFLT_PREVENT_STARVATION_ENABLED Default flag for preventing starvation of lower priority jobs.
static int	DFLT_PRIORITY Default priority that will be assigned if job does not have a priority attribute set (value is 0).
static String	DFLT_PRIORITY_ATTRIBUTE_KEY Default priority attribute key (value is grid.task.priority).
static int	DFLT_STARVATION_INCREMENT Default value on which job priority will be increased every time when job gets bumped down.
static int	DFLT_WAIT_JOBS_NUM Default waiting jobs number.

Fields inherited from class org.apache.ignite.spi.IgniteSpiAdapter

ignite, igniteInstanceName

Constructor Summary

Constructors

Constructor and Description
PriorityQueueCollisionSpi()

Method Summary

Modifier and Type	Method and Description
protected List<String>	getConsistentAttributeNames() Returns back a list of attributes that should be consistent for this SPI.
int	getCurrentActiveJobsNumber() Gets current number of jobs that are active, i.e.
int	getCurrentHeldJobsNumber() Gets number of currently 'held' jobs.
int	getCurrentRunningJobsNumber()
int	getCurrentWaitJobsNumber() Gets current number of jobs that wait for the execution.
int	getDefaultPriority() Gets default priority to use if a job does not have priority attribute set.
String	getJobPriorityAttributeKey() Gets key name of job priority attribute.
Map<String,Object>	getNodeAttributes() This method is called before SPI starts (before method IgniteSpi.spiStart(String) is called).
int	getParallelJobsNumber() Gets number of jobs that can be executed in parallel.
String	getPriorityAttributeKey() Gets key name of task priority attribute.
int	getStarvationIncrement() Gets value to increment job priority by every time a lower priority job gets behind a higher priority job.
int	getWaitingJobsNumber() Maximum number of jobs that are allowed to wait in waiting queue.
boolean	isStarvationPreventionEnabled() Gets flag indicating whether job starvation prevention is enabled.
void	onCollision(CollisionContext ctx) This is a callback called: new grid job arrived executing job finished its execution topology changed periodically (on EventType.EVT_NODE_METRICS_UPDATED) When new job arrives it is added to the end of the wait list and this method is called.
PriorityQueueCollisionSpi	setDefaultPriority(int priority) Sets default priority to use if a job does not have priority attribute set.
void	setExternalCollisionListener(CollisionExternalListener lsnr) Listener to be set for notification of external collision events (e.g. job stealing).
PriorityQueueCollisionSpi	setJobPriorityAttributeKey(String jobPriAttrKey) Sets job priority attribute key.
PriorityQueueCollisionSpi	setName(String name) Sets SPI name.
PriorityQueueCollisionSpi	setParallelJobsNumber(int parallelJobsNum) Sets number of jobs that can be executed in parallel.
PriorityQueueCollisionSpi	setPriorityAttributeKey(String taskPriAttrKey) Sets task priority attribute key.
PriorityQueueCollisionSpi	setStarvationIncrement(int starvationInc) Sets value to increment job priority by every time a lower priority job gets behind a higher priority job.
PriorityQueueCollisionSpi	setStarvationPreventionEnabled(boolean preventStarvation) Sets flag indicating whether job starvation prevention is enabled.
PriorityQueueCollisionSpi	setWaitingJobsNumber(int waitJobsNum) Maximum number of jobs that are allowed to wait in waiting queue.
void	spiStart(String igniteInstanceName) This method is called to start SPI.
void	spiStop() This method is called to stop SPI.
String	toString()

Methods inherited from class org.apache.ignite.spi.IgniteSpiAdapter

addTimeoutObject, assertParameter, checkConfigurationConsistency0, clientFailureDetectionTimeout, configInfo, createSpiAttributeName, failureDetectionTimeout, failureDetectionTimeoutEnabled, failureDetectionTimeoutEnabled, getExceptionRegistry, getLocalNode, getName, getSpiContext, ignite, initFailureDetectionTimeout, injectables, injectResources, isNodeStopping, onBeforeStart, onClientDisconnected, onClientReconnected, onContextDestroyed, onContextDestroyed0, onContextInitialized, onContextInitialized0, registerMBean, removeTimeoutObject, started, startInfo, startStopwatch, stopInfo, unregisterMBean

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.ignite.spi.IgniteSpi

getName, onClientDisconnected, onClientReconnected, onContextDestroyed, onContextInitialized

Field Detail

DFLT_PARALLEL_JOBS_NUM

public static final int DFLT_PARALLEL_JOBS_NUM

Default number of parallel jobs allowed (set to number of cores times 2).

DFLT_WAIT_JOBS_NUM

public static final int DFLT_WAIT_JOBS_NUM

Default waiting jobs number. If number of waiting jobs exceed this number, jobs will be rejected. Default value is Integer.MAX_VALUE.

See Also:

Constant Field Values

DFLT_PRIORITY_ATTRIBUTE_KEY

public static final String DFLT_PRIORITY_ATTRIBUTE_KEY

Default priority attribute key (value is grid.task.priority).

See Also:

Constant Field Values

DFLT_JOB_PRIORITY_ATTRIBUTE_KEY

public static final String DFLT_JOB_PRIORITY_ATTRIBUTE_KEY

Default job priority attribute key (value is grid.job.priority).

See Also:

Constant Field Values

DFLT_PRIORITY

public static final int DFLT_PRIORITY

Default priority that will be assigned if job does not have a priority attribute set (value is 0).

See Also:

Constant Field Values

DFLT_STARVATION_INCREMENT

public static final int DFLT_STARVATION_INCREMENT

Default value on which job priority will be increased every time when job gets bumped down.

See Also:

Constant Field Values

DFLT_PREVENT_STARVATION_ENABLED

public static final boolean DFLT_PREVENT_STARVATION_ENABLED

Default flag for preventing starvation of lower priority jobs.

See Also:

Constant Field Values

Constructor Detail

PriorityQueueCollisionSpi

public PriorityQueueCollisionSpi()

Method Detail

getParallelJobsNumber

public int getParallelJobsNumber()

Gets number of jobs that can be executed in parallel.

Returns:

Number of jobs that can be executed in parallel.

setParallelJobsNumber

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setParallelJobsNumber(int parallelJobsNum)

Sets number of jobs that can be executed in parallel.

Parameters:

parallelJobsNum - Parallel jobs number.

Returns:

this for chaining.

getWaitingJobsNumber

public int getWaitingJobsNumber()

Maximum number of jobs that are allowed to wait in waiting queue. If number of waiting jobs ever exceeds this number, excessive jobs will be rejected.

Returns:

Maximum allowed number of waiting jobs.

setWaitingJobsNumber

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setWaitingJobsNumber(int waitJobsNum)

Maximum number of jobs that are allowed to wait in waiting queue. If number of waiting jobs ever exceeds this number, excessive jobs will be rejected.

Parameters:

waitJobsNum - Maximium jobs number.

Returns:

this for chaining.

getCurrentWaitJobsNumber

public int getCurrentWaitJobsNumber()

Gets current number of jobs that wait for the execution.

Returns:

Number of jobs that wait for execution.

getCurrentActiveJobsNumber

public int getCurrentActiveJobsNumber()

Gets current number of jobs that are active, i.e. 'running + held' jobs.

Returns:

Number of active jobs.

getCurrentRunningJobsNumber

public int getCurrentRunningJobsNumber()

getCurrentHeldJobsNumber

public int getCurrentHeldJobsNumber()

Gets number of currently 'held' jobs.

Returns:

Number of currently 'held' jobs.

setPriorityAttributeKey

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setPriorityAttributeKey(String taskPriAttrKey)

Sets task priority attribute key. This key will be used to look up task priorities from task context (see ComputeTaskSession.getAttribute(Object)).

If not provided, default value is {@link #DFLT_PRIORITY_ATTRIBUTE_KEY}.

Parameters:

taskPriAttrKey - Priority session attribute key.

Returns:

this for chaining.

setJobPriorityAttributeKey

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setJobPriorityAttributeKey(String jobPriAttrKey)

Sets job priority attribute key. This key will be used to look up job priorities from job context (see ComputeJobContext.getAttribute(Object)).

If not provided, default value is {@link #DFLT_JOB_PRIORITY_ATTRIBUTE_KEY}.

Parameters:

jobPriAttrKey - Job priority attribute key.

Returns:

this for chaining.

getPriorityAttributeKey

public String getPriorityAttributeKey()

Gets key name of task priority attribute.

Returns:

Key name of task priority attribute.

getJobPriorityAttributeKey

public String getJobPriorityAttributeKey()

Gets key name of job priority attribute.

Returns:

Key name of job priority attribute.

getDefaultPriority

public int getDefaultPriority()

Gets default priority to use if a job does not have priority attribute set.

Returns:

Default priority to use if a task does not have priority attribute set.

setDefaultPriority

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setDefaultPriority(int priority)

Sets default priority to use if a job does not have priority attribute set.

Parameters:

priority - default priority.

Returns:

this for chaining.

getStarvationIncrement

public int getStarvationIncrement()

Gets value to increment job priority by every time a lower priority job gets behind a higher priority job.

Returns:

Value to increment job priority by every time a lower priority job gets behind a higher priority job.

setStarvationIncrement

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setStarvationIncrement(int starvationInc)

Sets value to increment job priority by every time a lower priority job gets behind a higher priority job.

Parameters:

starvationInc - Increment value.

Returns:

this for chaining.

isStarvationPreventionEnabled

public boolean isStarvationPreventionEnabled()

Gets flag indicating whether job starvation prevention is enabled.

Returns:

Flag indicating whether job starvation prevention is enabled.

setStarvationPreventionEnabled

@IgniteSpiConfiguration(optional=true)
public PriorityQueueCollisionSpi setStarvationPreventionEnabled(boolean preventStarvation)

Sets flag indicating whether job starvation prevention is enabled.

Parameters:

preventStarvation - Flag indicating whether job starvation prevention is enabled.

Returns:

this for chaining.

getNodeAttributes

public Map<String,Object> getNodeAttributes()
                                     throws IgniteSpiException

This method is called before SPI starts (before method IgniteSpi.spiStart(String) is called). It allows SPI implementation to add attributes to a local node. Kernal collects these attributes from all SPI implementations loaded up and then passes it to discovery SPI so that they can be exchanged with other nodes.

Specified by:

getNodeAttributes in interface IgniteSpi

Overrides:

getNodeAttributes in class IgniteSpiAdapter

Returns:

Map of local node attributes this SPI wants to add.

Throws:

IgniteSpiException - Throws in case of any error.

spiStart

public void spiStart(String igniteInstanceName)
              throws IgniteSpiException

This method is called to start SPI. After this method returns successfully kernel assumes that SPI is fully operational.

Specified by:

spiStart in interface IgniteSpi

Parameters:

igniteInstanceName - Name of Ignite instance this SPI is being started for (null for default Ignite instance).

Throws:

IgniteSpiException - Throws in case of any error during SPI start.

spiStop

public void spiStop()
             throws IgniteSpiException

This method is called to stop SPI. After this method returns kernel assumes that this SPI is finished and all resources acquired by it are released.

Note that this method can be called at any point including during recovery of failed start. It should make no assumptions on what state SPI will be in when this method is called.

Specified by:

spiStop in interface IgniteSpi

Throws:

IgniteSpiException - Thrown in case of any error during SPI stop.

setExternalCollisionListener

public void setExternalCollisionListener(CollisionExternalListener lsnr)

Listener to be set for notification of external collision events (e.g. job stealing). Once grid receives such notification, it will immediately invoke collision SPI.

Ignite uses this listener to enable job stealing from overloaded to underloaded nodes. However, you can also utilize it, for instance, to provide time based collision resolution. To achieve this, you most likely would mark some job by setting a certain attribute in job context (see ComputeJobContext) for a job that requires time-based scheduling and set some timer in your SPI implementation that would wake up after a certain period of time. Once this period is reached, you would notify this listener that a collision resolution should take place. Then inside of your collision resolution logic, you would find the marked waiting job and activate it.

Note that most collision SPI's might not have external collisions. In that case, they should simply ignore this method and do nothing when listener is set.

Specified by:

setExternalCollisionListener in interface CollisionSpi

Parameters:

lsnr - Listener for external collision events (null to unset the listener).

onCollision

public void onCollision(CollisionContext ctx)

This is a callback called:

• new grid job arrived
• executing job finished its execution
• topology changed
• periodically (on EventType.EVT_NODE_METRICS_UPDATED)

When new job arrives it is added to the end of the wait list and this method is called. When job finished its execution, it is removed from the active list and this method is called (i.e., when grid job is finished it will not appear in any list in collision resolution).

Implementation of this method should act on all lists, each of which contains collision job contexts that define a set of operations available during collision resolution. Refer to CollisionContext and CollisionJobContext documentation for more information.

Specified by:

onCollision in interface CollisionSpi

Parameters:

ctx - Collision context which contains all collision lists.

getConsistentAttributeNames

protected List<String> getConsistentAttributeNames()

Returns back a list of attributes that should be consistent for this SPI. Consistency means that remote node has to have the same attribute with the same value.

Overrides:

getConsistentAttributeNames in class IgniteSpiAdapter

Returns:

List or attribute names.

setName

public PriorityQueueCollisionSpi setName(String name)

Sets SPI name.

Overrides:

setName in class IgniteSpiAdapter

Parameters:

name - SPI name.

Returns:

this for chaining.

toString

public String toString()

Overrides:

toString in class Object