See the WebSphere eXtreme Scale Wiki for links to eXtreme Scale Version 7.0 documentation.
If you log in
with your developerWorks ID, you can leave comments and feedback for the development team.
ObjectGrid has the extreme transaction processing capabilities.
ObjectGrid is a grid-enabled, memory database for applications that are written in Java. ObjectGrid can be used simply as an in-memory database, or it can be used to distribute data across a network. Persistence is implemented through memory-to-memory replication, which can be configured to be synchronous, asynchronous, or a combination of the two. The ObjectGrid does not directly push state to disk, but you can push data to disk by using vendor-acquired software plug-ins. ObjectGrid provides vertical scalability by judicious use of threading. You can accomplish horizontal scalability by partitioning the data. A partition is synonymous to a hash column, where all data that is hashed to the specific column is hosted in a specific partition. A partition is made up of exactly one primary and a configurable number of replicas. These pieces of a partition are called shards. A single Java Virtual Machine (JVM) might host many partitions, but never more than one shard from the same partition. Very large sets of data can be stored in memory this way across large numbers of JVMs.
Self-contained grid
ObjectGrid is packaged as several Java archive (JAR) files. When you are running ObjectGrid by itself, the objectgrid.jar file is generally all that is required. You can use additional JAR files for prerequisite utilities like the cglib.jar file that may also be required. Optionally, the ogstreamquery.jar file is available to use the stream query support.
Self-organizing grid
ObjectGrid organizes partitions across the network of available JVMs based on the policies that are specified in the deploymentPolicy.xml file. The deployment policy is described in detail in the Environment section. The deployment policy is a policy file that is used to describe the ObjectGrid behavior. The primary of a partition can never be co-located in a JVM with one of its replicas. The primary and replica are also not allowed to share a network interface card (NIC), unless that sharing is allowed by the policy. Zone support allows more sophisticated configurations to control placement across data centers. With this capability, grids of thousands of partitions can be easily managed using a handful of optional placement rules.
Self-repairing grid
With the self-organizing ability, ObjectGrid can also self-repair. As servers leave the grid, ObjectGrid tries to maintain the replica count. If a server that is holding a shard for a partition fails, then ObjectGrid tries to recreate the shard on a surviving server automatically if possible. If the application specified a primary and two replicas for each partition, then ObjectGrid automatically tries to maintain the primary and replica count as failures occur. The high availability manager in WebSphere Application Server provides this failure detection. This high availability is available whether ObjectGrid is used with WebSphere Application Server, or in standalone mode.
© Copyright IBM Corporation 2007,2009. All Rights Reserved.
