Font Size

SCREEN

Mirrored Storage

Our mirrored storage engine fits in the space between SA Forum checkpoints (mirrored memory area) and traditional IMDBs (in memory databases). It is an API for generic persistent storage and active/standby mirroring of application data. High throughput, low CPU utilization and low latency are key features, as well as working well together with high availability middlewares and associated fault and recovery scenarios.

The data is arranged as objects, actually user-defined structures directly in C/C++, which may have links to other objects and allocated members. These objects can be inserted, modified and removed when in active state, and if there is a standby instance it will get the corresponding updates, with user callbacks if desired.

 

It was designed specifically with stateful network communications software in mind, where simple mirrored RAM is too simplistic and fully-fledged databases may be too slow, costly and unnecessarily complex. It was also designed for zero-maintenance and 24/7 workload with low response times at all times. This means that there is no blocking checkpoint generation (and no CPU utilization peaks), and a fresh standby can be synchronized in runtime without blocking the active instance.

Updated: Tuesday, 08 February 2011 13:44

Home SDKs Mirrored Storage