Special Features:
SUN GRID CASE STUDY: COGNIGEN
CORPORATION
"By implementing the Sun Grid System and Sun's Netra X1 servers, we have
drastically lowered the total cost of ownership and dramatically increased
productivity. Since implementation, each one of our scientists has one extra
hour every day to work on other things. We are getting more work done with the
available resources."
Duncan Ross
Director of Information Technology
Cognigen Corporation
CASE STUDY SUMMARY
Company: Cognigen Corporation
Area of Focus: Cognitive engineering - medical research
Industry: Healthcare
- Sun Products:
- Sun Netra X1 servers
Sun Enterprise 450 servers
Solaris Operating Environment
Sun ONE Grid Engine software
Application: Data Analysis
Cognigen Corporation provides data analysis and consulting services for the
pharmaceutical, biotechnology, and healthcare industries. The company combines
innovative software tools with advanced Internet technology, statistical
analysis, and simulation methodology, and human creativity to quickly and
efficiently assemble data, perform analyses, and communicate results to global
teams for knowledge-based decision making. Cognigen calls this process
"cognitive engineering," and it provides the principles and discipline to
maximize human, product, and company potential.
Headquartered in Buffalo, New York, Cognigen is growing rapidly. Since
1992,
many of the leading pharmaceutical companies have partnered with Cognigen to
develop products that are powered for success. Cognigen needed to deploy
sufficient resources to meet growing workload demands. To accommodate their
needs and rapid growth, Cognigen turned to the Sun Grid Engine software from
Sun Microsystems Inc. to provide dependable, consistent and inexpensive access
to computerized capabilities.
Practices of the Past Cognigen's scientists use a highly CPU-intensive
application, NONMEM, which is architecture and compiler dependent. NONMEM is
the core of Cognigen's business as it allows modeling for variability within a
single patient and across a population - predicting the effect of the drug on
a given population. NONMEM does not take advantage of multithreading or
multiprocessing, and job run times vary anywhere from hours to days.
In the past, Cognigen's mechanism for running these jobs was simply to log
in
to one of their Sun Enterprise 450 servers, check how many NONMEM jobs were
running, and submit the job. If all the system processors were used (four
NONMEM jobs running), Cognigen's scientist would log in to another machine and
check the number of NONMEM jobs running on that machine. If all computers were
at capacity, the scientist would have to check again later to see if a CPU
became available.
"This process was inefficient in many ways," says Duncan Ross, Director of
Information Technology at Cognigen. "It's a classic growth problem - In the
early days of the company, this really wasn't an issue as there were only a
few scientists. As the company continued to grow rapidly, more users were
running increasingly complex jobs. The process of checking for available CPUs
became extremely taxing on the end user. Clearly, in order for users to remain
productive, they had to focus on design and development, rather than hunting
for resources."
The Search for a New Solution Cognigen had several key criteria when
selecting
a new Information Technology (IT) solution. "One of the top challenges was the
need to handle dynamically changing workloads. So flexibility is key," adds
Ross. Cognigen needed to find a mechanism for submitting jobs from a central
location and the chosen tools needed to be portable to other systems. The
company preferred Free/Open Source tools, while utilizing less expensive and
faster hardware. Lastly, Cognigen had standardized on SPARC architecture for
NONMEM job processing, so any and all new systems had to be of the same
architecture.
Sun Grid Makes the Grade After evaluating a number of solutions, Cognigen
chose the Sun ONE Grid Engine software and Sun's Netra X1 server. The Sun ONE
Grid Engine software enables Cognigen to submit and manage jobs from just
about any UNIX or Linux system on their network. The command line utility
gives the company the flexibility to script and automate jobs, as well as
build a custom front end. The GUI provides Cognigen with a convenient
management tool for administering the Grid. "By implementing the Sun Grid
System and Sun's Netra X1 server, we have drastically lowered the total cost
of ownership and dramatically increased productivity," says Ross. "We
experienced a 10 to one reduction in hardware costs and almost doubled the CPU
power. Since implementation, each one of our scientists has one extra hour
everyday to work on other things. We are getting more work done with the
available resources." At the start of Cognigen's project, Sun introduced the
Netra X1 Rack Mount Server. "Because NONMEM does not take advantage of
multithreaded SMP, the Netra X1 server was just the low cost solution we were
looking for," says Ross. "With this amount of horsepower, we can process
twelve times the amount of NONMEM jobs, allowing us to do more thorough and
complex analyses."
Primed for the Future Cognigen is well positioned for rapid growth with
Sun's
Grid Computing. Now, Cognigen can complete analyses much more quickly and
ensure their clients have better analysis results much earlier. With the Sun
Grid Computing Environment both new and existing systems can be combined to
deploy massive computing power wherever and whenever it's needed most.
Additionally, it allows Cognigen's end users to find resources quickly, use
them efficiently and scale them seamlessly. As Ross explains: "What is,
perhaps, most important is the capacity of the Sun Grid Computing can always
be increased by simply adding more computing power."
|
