RFP for Beowulf Linux Cluster for

High Performance Computing for

Information Technology and Communication

January 2002

 

 

1.      General goals for High Performance Computing at the University of Virginia:

1.1.   Maintain a level of institutional computational capability that is consistent with peer institutions.

1.2.   Provide platforms for high performance computing in support of the research activities of U.Va. faculty that require more resources than can be accommodated with a single user's workstation

1.3.   Identify and maintain an upgrade path that is consistent with national trends in High Performance Computing, and which would be scalable in the event that new funding opportunities occur.

1.4.   Provide appropriate platforms for development and testing of applications that may later migrate to national supercomputer centers for production work.  The systems ought to be powerful enough to demonstrate capability and increase the likelihood for success in applying for time at national supercomputer centers.

1.5.   Provide the appropriate facilities that will permit researchers, including both graduate and undergraduate students, to gain necessary experience with parallel systems and parallel programming paradigms.

1.6.   Provide platforms for University staff to develop expertise in the administration and uses of highly parallel systems.

1.7.   Move from outdated IBM SP2 with 24 nodes to Linux Beowulf cluster.

 

2.      Specific hardware needs relevant to these general high performance computing goals

2.1.   A system for developing and running large distributed memory parallel codes.

2.2.   A system for developing and running shared memory parallel codes.

2.3.   A system for running many serial jobs. The performance of the system will be defined by how fast serial jobs can be executed and by the number of jobs a user can have running concurrently.

2.4.   Cluster nodes should be expandable and upgradeable.

 

3.      General System Requirements

3.1.   Red Had Linux, current distribution version

3.2.   Will purchase between 64 and 128 node cluster now with number of nodes depending on funding with an additional 48 to 128 nodes after July 1, 2002, with exact number of nodes depending on funding.

 

4.      Technical Requirements

4.1.   Mechanical Specifications

4.1.1.      Rack Mount

4.1.2.      As close to 1-U form factor as possible is preferred, 1-U form factor is not required.  No tower cases.

4.1.3.      Prefer at least one available (empty) PCI slot

4.1.4.      Prefer space for two internal disk drives

4.1.5.      Provide sufficient Uninterruptible Power Supplies (UPSes) of at least 3000 Kva UPS for each 16 nodes.

4.1.6.      Provide all needed mounting hardware, racks, slides, etc.

4.1.7.      Temperature inside the rack with door closed should be ambient room temperature.

 

4.2.   Electronics

4.2.1.      High-end Proposal

4.2.1.1.            Compute Nodes

4.2.1.1.1.                  Dual-P4 Xeon (Intel Pentium) processors preferred, AMD Athlon processors/chipsets acceptable.

4.2.1.1.2.                  At least 1 GB RAM, either RDRAM or DDRAM.

4.2.1.1.3.                  Copper 1 Gbps Ethernet network interface preferred, other interfaces (Dolphin-net, Myrinet, etc) considered.

4.2.1.1.4.                  IDE hard drive preferred.

4.2.1.1.5.                  At least one free PCI slot of at least 64bit/33Mhz.

4.2.1.1.6.                  KVM with full cabling.

4.2.1.2.             Server Node (one)

4.2.1.2.1.                   Dual-P4 Xeon (Intel Pentium) processors preferred, AMD Athlon processors/chipsets acceptable.

4.2.1.2.2.                  At least 2 GB RAM, either RDRAM or DDRAM.

4.2.1.2.3.                  Copper 1 Gbps Ethernet network interface preferred, other interfaces (Dolphin-net, Myrinet, etc) considered.

4.2.1.2.4.                  Integrated 10/100 copper Ethernet interface

4.2.1.2.5.                  Two 65Mbit/66Mhz PCI buses

4.2.1.2.6.                  RAID controller

4.2.1.2.7.                  At least 200 GB of level 5 RAID fast SCSI storage

4.2.1.2.8.                  Keyboard, mouse, CD-ROM, and monitor required

 

4.2.2.      Low-end Proposal

4.2.2.1.            Compute Nodes

4.2.2.1.1.                  Dual-processors

4.2.2.1.2.                  At least 1 GB RAM, either RDRAM or DDRAM.

4.2.2.1.3.                  Integrated 10/100 copper Ethernet interface preferred, other interfaces considered.

4.2.2.1.4.                  IDE hard drive preferred.

4.2.2.1.5.                  At least one free PCI slot of at least 64bit/33Mhz.

4.2.2.1.6.                  KVM with full cabling.

4.2.2.2.             Server Node (one)

4.2.2.2.1.                   Dual processors

4.2.2.2.2.                  At least 2 GB RAM, either RDRAM or DDRAM.

4.2.2.2.3.                  Integrated 10/100 copper Ethernet interface.

4.2.2.2.4.                  One 65Mbit/66Mhz PCI buses

4.2.2.2.5.                  RAID controller

4.2.2.2.6.                  At least 200 GB of level 5 RAID fast SCSI storage

4.2.2.2.7.                  Keyboard, mouse, CD-ROM, and monitor required

 

5.      Software

5.1.   Current Red Hat Linux distribution required for this Beowulf Red Hat Linux Cluster

5.2.   Propose firm’s available high performance computing software, to include, but not limited to:

5.2.1.      Cluster management

5.2.2.      Compilers – Portland Group Compilers with IMSL libraries

5.2.3.      Libraries: MPI, etc

 

6.      Integration Services

6.1.   Propose full turnkey hardware installation services for the Beowulf cluster. 

 

7.      Contents of Proposal

7.1.   Technical on proposed hardware

7.2.   Propose pricing for available configurations (high-end, low-end, or both). Include per-node pricing.

7.2.1.      64-node cluster

7.2.2.      128-node cluster

7.2.3.      Additional 48 to 128-nodes to be added to cluster for delivery and payment after July 1, 2002

7.2.4.      Hardware installation of all nodes in racks, cabled, switches, ready to run.

7.2.5.      After installation, all nodes up, running simultaneously for a 3 to 5 day burn-in test.

7.2.6.      What donations the firm plans to provide?

7.2.7.      Individual pricing for software components listed above

7.2.8.      Specify hardware warranty period

7.3.   Delivery of hardware

7.3.1.      Must be delivered no more than 30 days after contract

7.3.2.      Proposed equipment must be available for shipping by March 1

7.3.3.      Vendors must propose the penalty clause for late delivery

7.3.4.      May 1st installation deadline

 

8.      Vendors from whom to specifically solicit proposals.

8.1.   Dell Corporation Advanced Systems Group/PA1 Education, 804-763-3650

        13811 Turtle Hill Road, Midlothian, VA 23112

 

8.2.   Linux Networx, Inc.  1-877-505-LNXI or 801-562-1010

8689 South 700 West, Sandy, UT 84070      E-mail: info@linuxnetworx.com

 

8.3.   Microway, Inc    1-508-746-7341

Research Park, Box 79, Kingston, MA 02364   E-mail: info@microway.com