Monday, October 25, 2010

High Performance Computing and the Latest Applications

Last week, I keynoted the Harvard High Performance Computing Summit and updated the attendees on the latest research computing and novel applications at Harvard Medical School.

1.   Orchestra - is our 6000 core high performance computing infrastructure that is shared by all research departments.  They key to its success has been the sense of community we've developed around a shared utility.   The school funds the power/cooling/hosting/basic infrastructure as well as an expert staff to serve the faculty.   Researchers use their grants to add computing power and storage to the community utility instead of building small clusters under desks and in local mini-data centers.

2.   Novel Sources of Data  - The American Recovery and Reinvestment Act (ARRA) and its related healthcare IT regulations are encouraging clinicians and patients to record healthcare data in electronic form.   With proper privacy controls and institutional review board oversight, this new data can provide support for comparative effectiveness research, population health monitoring, and pharmacosurveillance.

3.   Mining those novel sources of data with Shrine - Petabytes of data does not change healthcare, but transforming that data into information, knowledge and wisdom does.   Shrine enables researchers, with appropriate oversight and approval,   to study the de-identified data  of 10 million patients at Harvard affiliates.

4.  Building collaborations with Profiles - The path to wisdom requires that researchers ask the right questions.   Assembling a team of colleagues into an optimal multidisciplinary combination is easy with our Profiles social networking application.   It mines the literature, Harvard directories, and the connections among all Harvard faculty to identify the right people to ask the right questions.

5.  Sustaining the research infrastructure and applications with an NIH compliant, researcher  friendly business model is challenging.   Computing power, data, tools, and collaborators are great but they must be supported with operating budgets, grant dollars, and philanthropy.   Harvard has retained an expert consultant and is moving forward on an aggressive schedule (60 days) to analyze all the potential sustainable business models (direct grant funding, indirect overhead, chargebacks, school operating budgets, transaction fees, subscription fees) that will support growth in infrastructure and staff over time.

The high performance computing summit was a great opportunity to share ideas and learn from each other.    I look forward to sharing the sustainable business model for research computing as it is finalized by the end of 2010.

1 comment:

Tomcat said...

In regard to Orchestra, Intel has a nice, little video of a proof-of-concept cooling scheme for data centres:

http://video.intel.com/?fr_story=2d6e0fbbef76b72c6119cc7fe7889bba20cb5192&rf=bm

Does Orchestra utilize free cooling, or, as do most data centres, does it consume about 1/3 of the power coming into the room to power air conditioners.