Finally a large-scale distributed key value fit for Exascale opportunities! Radon leaves existing distributed key-value systems in its wake by breaking the 1000+ node barrier and Exabyte data capacity limit while providing sustained performance. Critically important to its scaling capabilities, Radon distributes objects uniformly across a very large cluster, enabling high performance point operations, while also devising mechanisms to perform efficient range-query and ordered traversal of data. Radon's high-performance is fully leveraged to address the needs of applications demanding in-memory data access speeds at Flash/NVM memory persistency and operating costs.
Built on top of Levyx’s HeliumTM, one of the world’s fastest embedded key/value store solutions, Radon is highly optimized for ultra-low latency inserts, lookup and iterate operations, non-volatile memory (NVM) storage fabrics such as NVMe Flash and SCM and serves as a data-centric compute offload. Unlike Helium and other single node KVS object stores, Radon offers an extensive set of enterprise features including replication, elasticity, consistency, and support for transactions. In addition, Radon uses Levyx’s Xenon™, a distributed analytics platform, to offload compute and analytics and imposes a schema on key/value pairs to apply just-in-time compilation to run analytics on bare-metal. Finally, Radon is written entirely in C and is backed by 16 patents and patent applications, and its source code and IP are 100% owned by Levyx.
- Radon is a large-scale distributed key/value object store optimized for:
- Exabyte use cases @ 1000+ nodes scale
- Ultra low latency insert, lookup and iterate operations
- NVM storage fabrics such as NVMe Flash and Storage Class memory
- Data-centric compute offload using JIT bare-metal execution
- Radon is Enterprise grade, having features such as:
- Radon is suitable for both transactional and analytic workloads
How to Run It
Radon is built to run on a vast spectrum of platforms, both on-premise and in the cloud. The core requirement for Radon is the availability of fast storage and network layers. It can run directly on storage devices and can also use one or multiple files as its targeted data repository. Radon works on any storage abstraction product in the market (virtual disks, software or hardware raid, etc.) and can use memory directly as its underlying storage media.