ACADEMIA

Linux Labs, the world leader in third-generation Linux-based supercomputers, announced at the LinuxWorld trade show a key stepping stone towards large-scale deployment of the Linux(TM) Operating System. Called Clusgres(TM), this product has the ability to make standalone applications run in parallel across a group of computers and act as one, including the Open Source database system PostgreSQL. To date there has been no capability to either increase speed of free database servers for Linux or to allow them to scale past the limitations of a single server. Despite advances in clustered supercomputing, there has been no capability for a server program running on a single machine, let alone the PostgreSQL database, to transparently operate across linked computers and to operate as if a single large program. Now, Clusgres(TM), an implementation of Linux Labs' parallelization libraries called libOPUS(TM) or Open Parallel Unified System, permit any standalone server program to operate in parallel on a group of computers with no modification to the original program whatsoever. These libraries work closely with the SCI interconnect technology, an innovative switching circuit or "backplane" from Dolphin Interconnect Solutions. Dolphin's products provide a high-performance interconnect for multiprocessing. Special features that enable libOPUS(TM) performance are the low latency and efficient handling of small data communication. The Clusgres(TM) technology which creates cluster capability of PostgreSQL, an SQL-94 compliant database system, permits large amounts of queries to be sent to a database cluster and, instead of being processed in serial by one server (as if in one long line), they are distributed in parallel to every "compute node" and answered simultaneously. Using libOPUS(TM), one additional single-processor node to a cluster creates near double performance to a standalone database server. The Linux Labs system was programmed by their chief programmer, Suchindra Katageri, and runs in the low-level C language. Clusgres(TM) will be offered for both corporate and educational applications, and Linux Labs will be offering the system as a hardware/software combination, a co-located database service, and as ready-to-use cluster computers. These clusters are joined together using Linux Labs' Nimbus(TM) cluster operating system which makes all "nodes" in a group of computers appear as one large computer and permits "checkpointing," where tasks can be stopped and restarted at will. "Rarely has a set of reference libraries been able to benefit so many applications without them being completely reprogrammed. We are hoping that the emerging Linux industry can benefit by these tools for enterprise-level scaling of the upstart operating system," said Steven James, Director of Research for Linux Labs.