Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Message-Id: <v04020a0bb48ee581f897@[18.177.0.74]>
Date: Tue, 28 Dec 1999 17:19:09 -0500
To: infosys@mit.edu
From: Greg Anderson <ganderso@MIT.EDU>
Subject: Discovery Project Announcement: Longjobs

Everyone,

I'm pleased to announce the launch of an I/T Discovery project to explore
the ability to run long jobs on Athena without tying up public cluster
workstations. This project, commonly known as 'longjobs', has a _long_
history; the Discovery project represents an agreement among major
stakeholders in IS to explore the issue in a concentrated, project-focused
manner. The Discovery highlight will be agreement among IS and customers
regarding whether or not IS should provide such a service on an
enterprise-wide scale.

On Friday, Dec. 17, a variety of IS staff representing Academic Computing,
Faculty Liaisons, Athena Operations, Network Operations, Team Athena, and
ITLT Directors for  Discovery, Delivery, Integration, and Support, agreed
that a Discovery project should go forward. The Discovery team will check
in with this group again as more information is known about the current
needs. Earlier this month, selective, current information was gathered that
substantiated the ongoing need for a longjobs capability. One task for the
Discovery team will be to validate the needs more broadly in today's campus
computing environment.

The Discovery project focus on the following deliverables:

1. Business case. This is a major facet of the Discovery project. The longjobs
business case should be based upon historic and current needs expressed for
this service and an enterprise-wide assessment of the need and potential
value delivered. The Business case has multiple perspectives: customer
(students and faculty), IS staff (accounts, Athena cluster patrol, Stopit)
and policy perspective (Athena rules of use). Activities in this phase
include qualifying and quantifying the need with customers, estimating
steady state demand, and discussion about the appropriate business model.
Planning and fostering changes in user behavior will be an important part
of the business case.

2. Building upon the work that has already been done, the Discovery team will
focus on the technical design of the architecture. Of particular interest
will be integration with the Athena environment, scaling, robust
performance, and insuring that security is maintained.

3.  The team should address the proposed design and build budget model and
other costs for ongoing service and support. Discuss the administrative
issues related to long jobs - administering the service, establishing
service reporting/monitoring for future growth, providing support and
service activities, etc.

4. If customer response indicates the need to explore the service in greater
detail, the team will be asked to work to size the need (and insure
scalability), estimate the cost (both start-up and operational). If the
demand warrants, the team will be challenged to design a service that is
supportable and that meets normal demands but is also a service that has
appropriate controls, and provides incentives and disincentives for
appropriate use.

5. Determine the role of a pilot project in the work, identify the measures
that such a pilot might yield, and factor those data into the final
recommendations.

In the current Athena environment long-running jobs submitted by users mean
that the workstation is tied-up for the length of that job.  Athena rules
of use stipulate that machines cannot be unattended for longer than 20
minutes. Given the increasing demands for available Athena workstations,
especially at the end of the academic term, there has been a steady demand
for a long job service in the Athena environment.  Since 1994, faculty in
multiple courses (1, 3, 4, 6,10,13,15,16, 22) have asked for long jobs
capabilities and have assigned long-running work as part of their
curriculum. In addition, graduate students have an ongoing need to run
compute intensive jobs as part of their degree work.

Vijay Kumar is the sponsor for the project. It is identified in the IS
Strategic Plan to Deliver new I/T products, support and services and is a
component of the strategy to improve and extend the Athena academic
computing services and environment. Orginally, this project was projected
to begin later in the Strategic Planning period (through June 2001).
Because of the changing enviornment and to leverage resource availabilty
now, it was agreed to move the project forward.

Bob Basch is the project leader. He will devote 50% of his time to the
project. Other team members are Ted McCabe, Anne Salemme, and Abby Fox.
Other IS staff will be involved as needed. The team's mailing
list is longjobs@mit.edu and the project notebook is available at:
http://web.mit.edu/longjobs/

The team will report its findings to the key stakeholders by mid-March.

Please join me in supporting this project.
Greg