HPC System Acquisition and Service Provision

1
HPC System Acquisition and Service Provision
Request for Information

• José L. Muñoz
• Deputy Office Director
• Office of CyberInfrastructure
• (OCI)
• 9 September 2005

2
Agenda

• 900am Background Crawford
• 930am Introduction Muñoz
• 1000am Panel I Ahalt
• Acquisition Models
• 1100am BREAK ALL
• 1115am Panel II Heisey
• HPC Vendor Perspective
• 1215pm LUNCH (on your own) ALL
• 100pm Panel III Davis
• Benchmarking/Purchase Options
• 200pm Wrap-Up/Discussion ALL

3
Purpose of RFI

• Distribute information
• Gather information
• Answer questions
• Promote dialog between potential RPs and
  potential vendors
• Information is being recorded and will be made
  publicly available
• Please identify yourself when speaking

4
Registered Demographics

• Universities
• Case Western Reserve U.
• Cornell Univ.
• Georgia Inst. Of Technology
• Indiana University
• Louisiana State Univ.
• NCSA
• Ohio Supercomputer Center
• Purdue
• PSC
• SDSC
• TACC
• Univ. of NC
• Univ. of Utah
• USC

• Vendors
• Cray
• DELL
• Hewlett Packard
• IBM
• Intel
• Linux Networx
• Rackable Systems
• SGI
• Sun Microsystems
• Other
• Argonne National Lab
• CASC
• DOD HPCMP
• Hayes Consulting
• NCAR
• ORNL
• Raytheon

5
Background

• NSF anticipates releasing one or more
  solicitations for the acquisition of
  high-performance computing (HPC) systems and
  support of subsequent HPC services. 
• Prior to the release of the solicitation(s), NSF
  invites interested parties to provide input on
  possible models for the acquisition and service
  support process. 
• Providing such input will not affect an
  organization's eligibility to respond to any
  subsequent solicitation(s).

http//www.nsf.gov/div/index.jsp?divOCI
6
Background

• NSF intends that its acquisition and support of
  HPC systems and services will be driven by the
  requirements of the science and engineering
  communities
• including their most computationally challenging
  needs. 
• Science and engineering communities will be
  involved in defining the requirements that guide
  NSFs HPC investments. 
• Acquisition metrics will include levels of
  performance on benchmark codes that represent the
  types of problems for which the system will be
  used. 
• The research and education community must have
  access to a stable and robust production quality
  HPC environment
• hardware reliability and the dependability of
  system software will also be factors in the
  selection process.

7
Acquisition Context

• One or more HPC systems will be acquired from one
  or more hardware vendors with funds supplied by
  NSF. 
• One or more resource providers (RPs) will manage
  and operate the system(s), including providing
  basic user support and services.

8
Possible Acquisition Models

• 1. NSF conducts an open competition to identify
  an RP or RPs.  Proposals include a description
  of a competitive process that the RP will use to
  acquire a system with funds provided by NSF. 
  Once RPs have been identified, each executes the
  process it proposed to select the system and the
  vendor from whom the system will be acquired. 
  The final award to an RP will include funds for
  the acquisition, through a sub-contract to the
  vendor, of the system identified.
• 2. NSF conducts an open competition inviting
  proposals from RP-vendor partnerships for the
  acquisition, management and operation of an HPC
  system.
• 3. NSF acquires a system directly from a vendor
  through an open competition and initiates a
  second competition for an RP site.

9
Other Topics for Discussion

• Metrics that might be used to define performance
  and reliability requirements in an acquisition
  process
• Selection criteria that might be used as a basis
  for proposal evaluation
• The strengths and weaknesses of acquiring an HPC
  system as a one-time purchase and/or of acquiring
  an HPC system in phases that meet a specified
  performance curve
• Strengths and weaknesses of alternatives such as
  leasing HPC systems.