RONNIEE Express Fabric - The In-Memory Network
The RONNIE Express In Memory Network unlocks an extremely compelling alternative to current design techniques using a cost-effective and disruptive architecture. The new In Memory Network is designed to remove all the limitations of today’s interconnection fabrics, providing military grade reliability. RONNIEE Express provides a fast and transparent way for any networked application to supercharge performance and acceleration. Remote memory accesses are transparent to the requesting processes by using RONNIEE, eliminating the need – and the bottleneck – of traditional RDMA support. There is no protocol stack involved in remote memory access, resulting in the lowest possible communication latency. How low? How about as low as 100s of nanoseconds for system to system latency and an orders of magnitude performance increase over conventional PCIe connections.
RONNIEE Express delivers the building blocks for high performance and scalable file system implementations along with petabyte-scale Big Data and HPC storage systems. This allows for extraordinary, unique and disruptive performance enhancements at the application and system levels, allowing applications to reach their full potential. Amplified performance that translates into enormous optimized operational and economic advantages in the enterprise.
A3CUBE’s RONNIE Express is an interconnection fabric using a unique combination of hardware and software. It combines multiple 20 or 40 Gbit/s channels with less than one microsecond latency, dramatically improving overall inter-system communication for advanced direct inter-server and inter-processor connections.
The RONNIEE Express Fabric simply uses an application-transparent mechanism, as the main communication model between two or more hosts. The advantage of this architecture is that once the memory mapping is defined, each active network interface has direct access to the In Memory Network without operating system intervention, resulting in the lowest possible communication latency, at any level, even for user-level software.