THE CORE MEMORY FORUM
April 23, 2017, 03:54:08 pm *
Welcome, Guest. Please login or register.

Login with username, password and session length
News:
 
   Home   Help Search Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: Celerity Computing  (Read 564 times)
JJW
Newbie
*
Posts: 7


« on: November 04, 2016, 06:21:36 pm »

Celerity Computing was not NCR but since it was formed by NCR employees (from Rancho Bernardo and Scripps Ranch in San Diego) and used an NCR processor chip, I thought a brief description is appropriate for this forum.

Celerity was formed in 1983 By Steve Vallender (President), Nick Aneshansley (V P of hardware development) and Drew McCrocklin (VP of software development) all from NCR. The idea behind the new company was that the high speed microprocessor that was used to execute Criterion firmware could be turned into a high speed, direct execution computer for graphic processing which would operate in an office environment and compete with other newly formed companies in an emerging marketplace (like Apollo and Sun Microsystems).

Other NCR employees were quickly brought on board: #4 was Karl Lehman (who would do device drivers) and #5 was me (J. J. Whelan, to work on the port of the Unix Operating System). Others on the software side were: Jeff Anderson (who was familiar with Unix and would also work on the port), Sandra Lee (to do utilities and who was just a general keep everyone's head focused sort of individual), Patricia Shanahan (who would port the Unix compilers), Clark Masters (to take care of operations, software tools and configuration management) and Stephanie McCartney (to do product builds and integration as well as general support). Hardware Engineers from NCR included Jim Kocol, Jim Gilbert and Gary Gilbert. On the software side we were joined by an excellent group of developers from the Burroughs office in Rancho Bernardo. The offices and final hardware assembly were located in a building in Scripps Ranch across the freeway from the NCR facility.

The main challenges were two fold: To turn the 16 bit architecture of the execution side of the chip into a full fledged 32 bit architecture and to provide virtual storage capability on a chip that didn't possess the fault mechanisms to properly suspend operations when a memory location was not available. Implementation was aided by the fact that the chip provided for a large number of external registers (intended for I/O) that could be used to implement the interface with the new hardware and (most importantly) that the chip had a full 32  bit access to external data memory. The founders had already come up with the basics for solving these difficulties and it fell to the rest of us to fill in the details.

This was an excellent team with both managers and others who had experience in designing, creating and delivering products to schedule. The initial development was done on a DEC VAX system and BSD Unix was the Unix base. The operating system goal was to exactly match the behavior of the VAX system.

The resulting system turned out to be more of a competitor with the DEC VAX products than with the originally intended graphics office systems.

The second system produced was a dual processing system and Jeff Anderson and I created what I believe was the first commercially available multi-processor Unix kernel.

The fact that the system was being used mostly in scientific processing environments led to the decision that we should focus on the low end of the supercomputer market. The next product was an up to eight processor high performance system based on a high speed clone of the NCR chip build out of generally available integrated circuits. However, sales of systems didn't make up for the cost of development and the company was merged with FPS Systems in Beaverton Oregon.

FPS was a very successful manufacturer of scientific array processors which were attached to DEC and IBM equipment. They were encountering decreasing sales because DEC and IBM were providing their own array processors (often integrated with their system). The idea was that the merged company could provide a high performance scientific system with array processing capabilities that would compete with DEC, if not IBM. This hope was buoyed by the fact that FPS had a successful sales team and record in the desired marketplaces. The Mainframe development remained in San Diego under the same management though Clark Masters took over from Steve Vallender..

It was also realized that the Celerity team would not be able to keep up with the rapidly growing high performance micro-chips and it was decided to move away from the NCR based architecture. After looking into the possibilities, the decision was to use the Sun Microsystems SPARC architecture. A high performance SPARC chip was already in development and Sun had no plans to develop a supercomputing system using it and that meant elerity would have Sun's full cooperation. SPARC also already had the, UCB Unix based, Solaris operating system and it was simple matter to move the multi-processing capabilities to that version.

The combined company sold several systems but was unable to maintain itself and was bought by Cray and then sold to Sun where it ended its life as the "Sun Supercomputing" division.
Logged
JimT
Newbie
*
Posts: 36


« Reply #1 on: December 06, 2016, 02:25:39 am »

Interesting history JJ.  I didn't know what happened to Celerity after everyone left.  I worked for Steve Vallendar, with Nick Aneshansley, and I worked with all of the hardware guys that went to Celerity.  Steve assembled a good group of people.
Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

Powered by SMF 1.1.4 | SMF © 2006-2007, Simple Machines LLC
Page created in 0.059 seconds with 16 queries.