Rai's Technology is what sets us apart. We've been on the leading edge of technology since the beginning of financial data integration. Applying modern design in the service of high performance and solving our customer's problems is what we live for.
Modern Design
The Rai Cache is built and configured around services. Each service works on its own, or in combination with other services, to provide the data-management, transformation, transport, protocol, clustering, and administration capabilities of the Rai Cache. The output of a service will often be the the input to another service. By cascading services within a single process, complex data enrichment is possible without incurring the cost of moving data between multiple processes. It is even possible to include feed handlers such as the Rai RDF Feed Handler in the Rai Cache to completely compress the data distribution tier if desired.
The Rai Cache's multi-threaded architecture allows the Rai Cache to fully utilize today's mutli-cpu and mult-icore hardware. So even with multiple services configured, the Rai Cache will typically be performance bound by the speed of the transport.
Performance
Here at Rai we've toyed with the idea of paying a third party to run some benchmarks in a lab, and then write a report that makes us look better than anyone else. But we like to think our customers are smarter than that.
So instead we rely on our customers to tell us how our products do in the real world. And what they've consistently told us is that we far outperform the competition in both latency and throughput.
Here's just a few reasons why:
- The Rai Cache is fully multi-threaded: we fully utilize a machine's resources.
- The Rai Cache uses TIBCO Rendezvous more intelligently. We've been doing this for a while, and we know how to get performance from Rendezvous that no one else can. In fact, fully loading a Rai Cache requires multiple RVDs.
- We're big fans of 64-bit Linux and we know how to squeeze every last bit of performance out the hardware we run on. Those annoying 40ms latency spikes you see from other vendors? We solved that long ago.
|