com.gigaspaces.grid.worker (GigaSpaces Technologies 6.0 Service Grid API Documentation)

Overview

Package

Class

Use

Tree

Deprecated

Index

Help

PREV PACKAGE NEXT PACKAGE

FRAMES NO FRAMES

Package com.gigaspaces.grid.worker

Provides ready-made GigaSpace Workers.

See:
Description

Interface Summary
GigaSpaceWorker	The GigaSpaceWorker interface defines the semantics of a worker entity that will process Entry objects from a GigaSpace.

Class Summary
AbstractSpaceWorker	The AsbtractSpaceWorker provides a framework upon which to build support for a GigaSpaceWorker.
TaskWorker	The TaskWorker provides an implementation of a GigaSpaceWorker that repeatedly takes Task objects from a JavaSpace and invokes the `Task.execute()` method.

Package com.gigaspaces.grid.worker Description

Provides ready-made GigaSpace Workers. Workers in the architecture implement a Command pattern, allowing the polymorphic processing of Task objects. Task Submission Management services create Entry objects, write the Entry objects to the GigaSpace and correlate results. Worker services are dynamically provisioned through the network to compute resources which meet operational requirements of the service.

Using the policy mechanisms available in the Service Grid, Worker services can have Service Level Agreements attached to them, ensuring operational criteria is met. By associating SLA Handlers with operational criterion is met. The dynamic grid worker's operational behavior can be monitored and dynamically scale its environment up and down, based on declarative attributes. In this way, the grid works with the Grid Service Management agents to dynamically size and provision the required services, based on realtime operational demands. An example of the SLA declaration in the deployment descriptor is shown here:

<SLA ID="backlog" Low="100" High="500" >
    <Configuration>
        <Component Name="org.jini.rio.qos.SLAPolicyHandler">
            <Parameter Name="slaPolicyHandler"
                Value="new org.jini.rio.qos.ScalingPolicyHandler(
                                             (org.jini.rio.core.SLA)$data)" />
        </Component>
        <Component Name="org.jini.rio.qos.ScalingPolicyHandler">
            <Parameter Name="MaxServices" Value="10" />
            <Parameter Name="LowerThresholdDampeningFactor" Value="1000" />
        </Component>
    </Configuration>
</SLA>

This declaration provides the definition of a SLA policy handler providing scalability, tracking GigaSpace backlog. If the backlog in the GigaSpace exceeds 500 entries, the scalable policy handler will trigger the creation of additional GigaSpace Worker instances to process the increasing amount of tasks, up to the limit of 10 (as specified by the MaxServices attribute). Conversely, if the backlog falls below 100, unneeded services will be scaled down.

The key to keeping the grid operating is to ensure grid services (and the tasks they are processing) are still running. Utilizing Fault Detection Policies, the Service Grid provides the essential capability to detect existence of running and available services. If services become unavailable, or fail to meet processing requirements, the recovery of those services to available compute resources is enabled by dynamic provisioning mechanisms. In an operational context these essential capabilities provide enhanced survivability and reconfigurability in such a way as to recover from damaged or failed resources and adapt to operational needs without a prior configuration. This fundamental capability provides support for the following:

Dynamic and adaptive (re) allocation of resources/services ensuring adequate load distribution of processing to the most appropriate resource based on declarative QoS requirements.
As demands to the system increase, processing will be distributed to available resources based on the ability for the resources to meet specified QoS requirements. As resources reach limits, new resources are added into the environment.
Distribution of processing across available hardware in a network-centric environment will be a key towards ensuring software performance. The effective distribution of processing based on increased demand will maximize the use of all hardware resources in the environment.

Configuration
The GigaSpace Worker supports the following configuration entry; where the configuration entry name is associated with the component name com.gigaspaces.grid.worker.

spaceWorker

Type:	`GigaSpaceWorker`
Default:	`TaskWorker`
Description:	This entry will be read at initialization by a GigaSpace Worker to determine the type of worker to instantiate.

templateClass

Type:	`Entry`
Default:	`Task`
Description:	This entry will be read at initialization by a GigaSpace Worker to determine the type of template to match when consuming work from the GigaSpaces.

Parameters
The GigaSpace Worker allows the declaration of parameters (name/value pairs), that control the behavior of the worker. The following parameters are provided:

Name	Default Value	Description
useTransactions	"no"	Whether transactions should be used when consuming work from the GigaSpaces.
useWeakHashMap	"yes"	Use of a `WeakHashMap` or a `HashMap` to store information processed by a `Task` Entry.
timeout	`Long.MAX_VALUE`	Timeout for the worker blocking take, if not defined the default is set to `Long.MAX_VALUE`.