GemFire SQLFabric is a horizontally partitioned memory-oriented SQL data management platform. It pools memory and disk across many nodes to offer:
- Predictable latency and very high throughput
Unlike disk oriented, centralized traditional database architectures, data structures and indexes are optimized for main-memory. The focus is on optimzing distributed data access rather than disk IO optimization.
- Dynamic and linear scalability
There is no need to ever over provision; simply add/remove data nodes to your cluster as and when demand pattern changes
When data access patterns is uniformly distributed, you can double the capacity to double database throughput or reduce the latency by half
- Continuous availability
every change is synchronously propagated to one or more "active" copies
or, even asynchronously to a remote cluster (multiple levels of failover)
there is nothing shared anywhere in the architecture
- Optimize by moving behavior to data
We introduce a novel concept called data aware procedures - an extension to standard stored procedures to allow the application to hint on the data the procedure is dependent on. The system can parallelize the execution of the data dependent behavior by moving the behavior to the data nodes. Results from each node can be streamed and aggregated.
See graphic below for a high level architecture