At the core, XAP implements a scalable, in-memory database which can be used as a data grid, a messaging grid, or as a parallel processing framework.
XAP makes it easy to scale the entire middleware layer (data, messaging, and services) of an application. It does this using an architecture which provides for just-in-time provisioning of processing resources, making it an ideal match for EC2. You can build and test an application on your laptop, and then migrate it to your own data center or to Amazon EC2 without any code changes.
The entire system runs under the control of an SLA-driven container. The container hosts applications, scales out to additional instances as needed, and manages partitioning, replication, and failover.
Applications can be built using C++, any .Net language, or Java via Spring, Hibernate, Tomcat, Mule, or J2EE. These applications can easily store plain (native) objects into the core storage facility provided by GigaSpaces.
It is easy to launch a GigaSpaces cluster on Amazon EC2 by following the directions in the tutorial. The platform is available on a per-hour basis, charged through Amazon DevPay, per their pricing schedule. You can also see the whole EC2 launch and management process in action in the screencast.
GigaSpaces will be presenting a pair of screencasts next month. On July 1st, they will talk about The GigaSpaces-RightScale One-Click-Cluster. On July 22nd they will talk about Scaling Applications on Amazon EC2 (I'll be participating in that one).