Main Page | See live article | Alphabetical index

Meiko Computing Surface

The Meiko Computing Surface is a distributed-memory massively parallel supercomputer with a variety of individual processors inside.

Meiko Scientific was founded by members of the design team working on the INMOS Transputer. When management suggested the release of the Transputer be delayed, Miles Chesney resigned and started work on machines based on the CPU. Nine weeks later they demonstrated a Transputer system at the SIGGRAPH in San Francisco in July 1985, which was released in 1986 as the Computing Surface.

The original Computing Surface (later known as the CS-1) used numbers of Transputer T414's connected together using their internal communications links, running a Unix-like OS known as MeikOS. The T414's were replaced with T800's as soon as these became available. Users were expected to write applications for the CS on other machines they owned, and then upload the programs for running on the machine. This process was improved with the addition of a standard Sun Microsystems workstation to act as a "front end". Users could then program using tools running under SunOS, which would be uploaded and run automatically.

As the speed of the Transputer fell in relation to the other RISC chips on the market (the competitive T9000 took too long to ship) Meiko added the ability to replace the T800's with Intel i860s. Systems were typically sold with one or two i860's on a card along with two T800's acting as a communications system (referred to as a Network Interface Processor). The cards also included 64kB of cache and up to 32MB of RAM. Sold as the Concerto or CS i860, the system was not very popular.

The final machine in the Meiko lineup was the CS-2, which replaced the i860's with SPARC CPU's. The expansion cards were similar to those in the earlier machines, mounting RAM, cache, and a SuperSPARC CPU, and could add an additional vector processing chip from Fujitsu. Unlike the earlier systems the CS-2 no longer used the T800 as a link, and instead replaced them with custom hardware that was much faster.

A major drawback of the Meiko architecture was poor I/O bandwidth for general data shuffling. Although aggregate bandwidth for special case data shuffling could be very high, the general case has very poor performance relative to the compute bandwidth. This made the Meiko computing surface uneconomic for many applications.

As of late 2002, the Meiko website was down, which appears to have been the case for some time.

The Meiko technical team was transferred to a joint venture company called Quadrics formed with Alenia Spazio in mid 1996. The Elan2 network ASIC (a directly attached UltraSPARC UPA device) was dropped and a new PCI based device (Elan3) was instead produced. The initial Elan3 platform was UltraSPARC based but was quickly replaced by Dec_Alpha based systems as Quadrics formed a new relationship with the Digital_Equipment_Corporation/Compaq. The combination of Quadrics and Alpha EV6 CPUs proved very successful and Digital/Compaq rapidly became one of the worlds largest suppliers of supercomputers. This culminated with the building the largest machine in the USA ; the 20TF ASCI Q supercomputer installed at Los_Alamos_National_Laboratory in 2002/2003. This machine consists of 2048 4*1.25GHz CPU EV67 Alphaserver SC ES45 nodes and two rails of the Quadrics QsNet network.

Quadrics also had success in selling Linux based systems, with the fatest Linux machine in the world (11.2TF) called MCR being installed at Lawrence_Livermore_National_Laboratory in 2002.

As of June 2003, 6 out of the top 10 fastest machines in the world were Quadrics connected.

External links: