The Bolo Page

Bolo is the name we've given to our systems. It is a Beowulf-class cluster, supercomputer. The name "Bolo" comes from the Keith Laumer series of books, where in the future, the Bolos are self-aware machines, that astonished their creators with their sense of unyielding honor. To see more, click here

There are many other computers of this nature that you can read about, in the link below. Most Beowulf supercomputers are built with commodity-off -the-shelf equipment (COTS), and many are using newer intel technology. Our cluster here was built by haperstance, accquiring most parts in various places during the past few years.   The systems are connected to high-speed internet connections, that also allows for remote computations. I am working on posting all the technical specifications and white papers will be posted here, in the near future.

And this is Gozer, our cat and supervisor of this project; here, she is building the systems, from the inside-out. To see some of the detailed specifications and a short FAQ, click here. Here is a tutorial presented by us, on August 5th, 2000 to the Linux Users Group, in the Washignton D.C. area

This is both of us, from the console part of the cluster...   This is Marie multitasking; if she could type with her toes, she would have a keyboard at her feet. Part of both clusters, upper and lower tier & part of the control console.

What  gives this system a different twist, is the heterogeneous enviroment; The lower tier is built from  SUN "sunrise" model sparc servers with external tower hard drives RAID's and external tower controller. They are in turn, connected to PC-based machines; AMD and pentium-based off-the-shelf hardware, with independent pentium front servers.       **** The links **** * The Beowulf project *Marie's 's page * Margret 's page *Gozer 's page

The upper pc-based cluster and the lower sun-based cluster allow for integration between both through software applications, and each cluster can run independent from each other. (We also have a third cluster) We run the upper tier for internet applications and the lower, for computation and load sharing, using MOSIX, and other parallel applications. The third cluster is win32-based architecture using RSH Services; The software allows clustering computers running Win32 with Unix and MPP's together into one parallel virtual machine.

Second cam shot. If you cant see it, click here

Connectivity

    OS support: kernel drivers (unix, solaris,Win32 bit)
    Maximum bandwidth: 1,000 Mb/s
    Minimum latency: 300 microseconds
    Acquired as: multiple-vendor hardware
    Interface port/bus used: ISA,PCI and VME bus
    Network structure: multiple hubs & routers
    Cost per machine connected: $2,500

Some VME busses are connected via SLIP; Although SLIP (Serial Line Interface Protocol) is usually considered to be at the low end of the performance spectrum, however, this allows some machines to perform socket communication via ordinary and custom high-speed RS232 serial ports.

    Linux and UNIX support: kernel drivers
    Maximum bandwidth: 0.1 Mb/s
    Minimum latency: 1,000 microseconds
    Available as: commodity hardware
    Interface port/bus used: RS232C & SCSI
    Network structure: cable between machines
    Cost per machine connected: $15.00

Presentation of this web page does NOT imply release of any information contained herein, to the public domain. All content, html,forms, and / or graphics presented here, are the sole property of MJG & MAT. Removal of this copyright notice is a crime under 17 U.S.C. 501. No part of this page or Web site may be used in whole or in part without our express written permission. Any material used on these pages that is not our property is used with permission of original author(s) or is in the public domain. MJG & MAT.copyright 1989-2000.