IBM WebSphere eXtreme Scale

IBM WebSphere eXtreme Scale is an IBM in-memory data grid product. It can be described as a fully elastic memory based storage grid. It virtualizes the free memory of a potentially large number of Java virtual machines and makes them behave like a single key addressable storage pool for application state. Applications can view this as a network attached storage medium. It is key addressable and applications can store a value at a key. Data within the grid can be replicated to achieve fault tolerance and protect against data loss. It can also be written to relational databases and any other persistence backend using write through or write behind technologies. Data can be automatically pulled from a backend such as a relational database or enterprise application if it isn't present in the grid.

IBM WebSphere eXtreme Scale can be used to access data from the grid significantly faster than from a traditional database. Applications which have a partitioned data model or data models that follow the constrained tree scheme approach will see linear scaling with products like this. IBM testing has shown that can be scaled out to hundreds of physical servers processing millions of transactions per second. Each server adds more CPU, memory capacity and network capacity to the grid for a linear scale up.

IBM WebSphere eXtreme Scale can be used with just a Java 2 SE JVM or can be used with an application server such as any version of IBM WebSphere Application Server from 6.0.2.x and higher.

It is marketed for extreme transaction processing. It is a complement to traditional transaction processing systems that have been available for decades in the form of high end SMP servers such as the IBM System z mainframe.

This technology allows very large memory stores to be realized. Terabytes of data can be kept in such a grid and then accessed from client applications that share the data within the grid. The data can be interacted with using the IBM WebSphere eXtreme Scales transactional client.