SUNet Backbone: Now at 10Gbit/s
The upgrade of the Stanford University network (SUNet) backbone to
use higher speed links, with a data rate of ten gigabits
per second, will be completed before the beginning
of Fall Quarter, 2007. During the past year, the equipment
in the backbone has been replaced or upgraded
to support the goal of having all major backbone links
running at this new standard speed. The number of routers
was also increased not only to spread the load, but
also to provide a second router for every network
to increase the reliability and availability of network
service. So while the standard speed of the connection
for an individual local network remains at one-gigabit
per second, the total capacity of the SUNet backbone system has been
increased ten fold.
Most local networks are now connected to one of the new "operational
zones." The initial topology has eight zones; six for the main
campus and two for student residences. The design provides
for dual routers for each local network to provide
for increased reliability along with the increased
capacity. Each zone has two switch/routers, which are
located in physically separate facilities to limit
the impact of a failure in any single facility. Each
switch also has a connected firewall to support local network migration
by the departmental Firewall project.
The firewalls are being operated
in their "high
availability" mode,
meaning only one is active at a time. The firewalls
get their network connections through their local zone
(as a switch, at the new standard ten gigabits per
second data rate), but they are otherwise operated
as independent routers for those networks which are
connected to them. The backbone is still composed of
two switch/routers located in diverse facilities. Each
operational zone has connections at ten gigabits per
second to each of the backbone routers.
Now that the backbone capacity upgrade is nearing completion, work
is underway to upgrade the link to the national research
networks by getting a ten gigabit per second link to
the high-performance research (HPR) network, part of
the CalREN networks run by the Corporation for Educational
Networks in California (CENIC). Stanford currently
has two one gigabit per second links to the CENIC HPR
network. Several researchers have been asking for bandwidth
in excess of one gigabit per second, so one of these
will be upgraded to run at ten gigabits per second,
like the SUNet backbone. Networking plans to have the
additional capacity available by the end of the 2007
calendar year.
- Mark Miyasaki
Shared Communication Services; Networking Systems
Research Computing Update
The term "research computing" spans
many topics. These range from using computers
to simulate experiments that might otherwise have been
done in a wind tunnel to computing with complex,
process-dependent applications. It also includes
aggregating large data sets from simulations and then
visualizing the results. The common factor is the need
for a massive amount of computational cycles.
Research computing has been taking place on the
Stanford campus for years.
What makes this an IT Services
issue is that the scale of many of these
early uses is increasing dramatically. There is a growth
in the number of faculty and researchers using, or
planning to use, large-scale research computing.
Traditional research labs
such as wet labs, social science survey labs, and mechanical
engineering air flow analysis are becoming astronomically
expensive to construct and require floor space that
just doesn't exist on campus. Small clusters of computers
have been used over the last few years to validate
the computer simulations. Many faculty are now ready
to make the jump to simulating large-scale problems.
To handle
the large-scale simulations, the 10–15 node clusters
currently used need to grow to 200–500
nodes per cluster. Current campus buildings are designed
to handle people, not systems. The 10–15 node
systems have worked so far, but have often put
the building at risk of blowing a breaker and
taking entire floors off-line. There are also very
demanding cooling issues for these systems.
IT Services has traditionally provided data center space
to handle the larger computing needs of the campus.
With the size of systems growing, and the number of
expected systems also growing, our late 1970's era
Forsythe Data Center isn't up to the task. The remediation
underway in Forsythe will help some, but is merely
the finger plugging the hole in the dike for the pent-up demand.
This is why Bill Clebsch has been focused on developing a new Data
Center strategy.
If we have a strategy and a plan available, we'll be
ready to implement it when the faculty clamor
for more computer space reaches the Provost.
IT Services is working with the
Dean of Research to insure we stay in sync with their
needs and directions. We are also working with SLAC.
The Committee on Academic Computing and Information
Systems (C-ACIS) will be undertaking a review
of research computing issues in the fall, and IT Services
will be in the middle of those discussions.
IT Services is also working to assess the viability of deploying
an application named Condor which would utilize the unused cycles
of the Linux cluster systems. These unused cycles would
be made available to campus faculty/researchers as
a start for large-scale research
computing. A "go-nogo" decision on this project is
expected in early August.
Two upcoming events will have IT Services support behind them. The
first is the third annual High
Performance Computing Day on campus. The event will be held
in conjunction with Mechanical Engineering at Wallenberg Hall all day
on Thursday, August 21.
Stanford and SLAC are involved in the Supercomputing
2007 conference in Reno, Nevada the week of November
12. This is an international event with many universities
and vendors participating.
- Phil Reese
Client Support; Account Management
Production Control Group Consolidation
As you may have seen in the last issue of its
in bits, Rick Lowe has
resigned his position as manager of the Production
Control Group (PCG). The current scope of the position
includes management of the Production Control
staff and oversight of the Satellite Operations Center
(SOC). The position has now been posted on StanfordJobs
and it will include the Business Continuity/Disaster
Recovery responsibilities as well.
With the expanded role, the new position will be at the Director level.
In the interim, I have hired a consultant, Gordon Spencer,
to fill the role of managing the PCG staff until the
position is filled.
This change will bring the 24-hour monitoring and support of the PCG,
the intra-group planning and coordination of the SOC,
and the preparation and drill processes of the Disaster
Preparedness team under a single umbrella. We believe
it makes strong organizational sense to consolidate
all of our emergency response functions.
With the planned retirement of
our current SOC leader in the next few years, this
is an opportune time to consolidate incident management,
disaster planning, and emergency response under the
auspices of the PCG.
The consolidated functions will continue to be called the Production
Control Group. Please feel free to direct any questions
to me or to Gordon.
- Jay Kohn
Shared Communications Services
Students Say Goodbye to PBN
On Thursday, July 26, IT Services reconfigured the phone switch to
no longer require a PBN (Personal Billing Number) for
domestic long-distance calls made by students. In our
efforts to simplify our customers' experience faculty,
staff, and students are no longer required to enter
a PBN when dialing domestic long-distance calls. International
calls still require the PBN, including calls to Canada
and Mexico.
Not long after the change, we
received our first HelpSU request from a student
inquiring if something was wrong (because they weren't
prompted for their PBN when making a long-distance
call).
- Christine Moe
Shared Communication Services; Communications Systems
Pathworks Upgrade
Our installation of Pathworks was upgraded
to version 2.5 on Saturday, July 21. Pathworks is the
tool that IT Services uses to hold procedural
guides and instructions, and it is used by
the staff in the Production Control Group, Service
Desk, and Order Processing to accurately execute daily
tasks.
The most notable benefit provided by this upgrade
is the ability to store any type of file within Pathworks.
Thus, it no longer requires that you convert your Word
or other files to PML (Pathworks-specific) files. This
release also supports Internet Explorer 7.
In the past
six to eight months, we have received several reports
of Pathworks procedures being cut off when
printed. That bug has been resolved in this upgrade.
There are also new import, export, and conversion functions,
export procedures to PDF and RTF/Word, and easier conversion
of Word documents to procedures. There are now scheduled
document reviews, a "subscription" feature that automatically
notifies users when a new version of a document
is published, a web editor, and automated review, approval,
and publication features.
Thanks to the team
that prepared for and executed this update: Kevin
Hall from UNIX Systems and Jose Rocha from Application
Support.
- Anne Pinkowski
Shared Communication Services; Application Support
The Stanford Sheriff...
...will be awarding a trophy at the upcoming Cart Parade.
Details in the next its in bits.
- The Cart Parade Committee