Multicore Computational Center
Multicore Computational Center
I am pleased to report that MC2 is continuing to show substantial progress this quarter towards achieving our MC2 mission, namely, "to evolve to a self sustaining leading national center in providing service oriented computational solutions employing multi-core computational technologies productively". Let me cite some of the ways we are gaining such recognition. In this respect, I am pleased to announce that two of our staff scientists, Dr.'s John Dorband and Yaacov Yesha received our first MC2 grant from the Laboratory of Telecommunication Sciences to work collaboratively with one of their scientists, Mike Ferguson, in developing a suffix array algorithm to run on our Cell cluster that can be used to compare strings consisting of at least 1 TB of data within 1 day of compute time. Working on this team is also our Research Assistant Nancy Walia, who prepared herself for this challenging assignment by using her course on Service Oriented Computing to do a project making use of 'bluegrit' to provide services for matching proteins based on a suffix array algorithm.
In addition, to meet the storage demands of the complex data intensive problems that utilize our 'bluegrit' cluster, such as our Service Oriented Atmospheric Radiance (SOAR) project, Google Inc. has stepped up and agreed to donate to the UMBC/MC2 a server with 2.2TB of high speed disk storage as a gift to support our immediate research needs. Moreover, Google indicated that they will host our 5 year collection of processed gridded IR thermal radiances from instruments on NASA's Aqua satellite and make these climate data sets available to the world community to study global warming as part of their cloud computing initiative they plan to launch this summer. Moreover, we are currently engaged in negotiating additional research projects with several other industries and agencies which I hope to announce in my next blog.
This is but one of the steps we at MC2 have initiated to become self sufficient and grow our reputation. Publications in journals and at conferences are another approach. Here too, we have been quite active having submitted over six papers during this quarter to prestigious journals and conferences. (See the publication tag to link to these papers.)
We are also quite mindful of our educational responsibilities to the UMBC institution and as a result we have initiated several programs to offer our students unique learning experiences. This spring we have offered for the first time a course by Prof.'s Yesha and Dorband titled "Introduction to Parallel Computing on the Cell Processor". MC2 purchased 11 Playstation 3s for this course and set them up in the same classroom as the course offered by Prof. Olano on Programming Games. For that course, 10 Xbox 360s purchased by the CSEE department are also installed in the same room. Class size was limited to 20 students who would share the 11 PS3s but close to double showed up to register or sit in and had to be turned away because of the limited seating available. As part of this course, they have invited several guest lecturers to describe their experiences implementing applications on a Cell cluster. In addition, Michael Perrone, IBM's Cell Application Manager also spoke to the class presenting them with an overview of the STK Cell simulator and how to program the SIMD processors. Thus, not only are students in both courses learning to program these systems but they are greatly enriching their learning experience through these guest lectures and class projects. In addition, due to the demand for SOA courses, Dr. Halem will again be offering his course this Fall his "Service Oriented Computing" class with emphasis on coupling the services to the 'bluegrit' computing infrastructure. The course also covers Globus for grid computing, and as a special topic for this coming year, he will be exploiting Hadoop open source software from the Apache organization for distributing and serving large files.
Over the intersession, IBM engineers came out to validate our Cell cluster software configuration and found our system to be conforming to the IBM standards thanks to our system administrator Zach Malone. They also agreed that two of our blades were flaky and put in a work order to have them replaced. However, as a recall of certain QS20s has been made by IBM, we expect to receive a complete replacement of our 12 Cell blades.
I also am pleased to announce that we were able to take on two more Research Assistants during this spring semester. They are Phuong Nguyen and Srinivasten Kannan. Phuong and David Chapman, are our two RA's working on the SOAR project and are doing a fantastic research effort in combining advanced concepts from aerospace physics and computational science in extending what originally was intended to be an on demand IR multi sensor service oriented grid processing system to a sophisticated climate process analysis system. Examples of the processes they are developing tools for tracking or mapping are such phenomena as the Madden-Julian Oscillations driven by the motion of deep convective clouds, El Nino /Southern Oscillation and long term trends of Outgoing Long Wave Radiation related to global warming. As noted in the earlier blog, we do intend in the future to discuss the implications, if not the details of some of our research projects. David and Phuong will be posting descriptions of their work under this site's research tag. We have also taken on an undergraduate, Brad Mancuso this spring, as a system admin in training under Zach. With the expertise of John Dorband overseeing the system and some expert users on our staff like David Chapman and Phuong Nguyen for application support, we are finally reaching a critical management level of cluster system admins to support the added work loads resulting from our collaborations.
As reported in my last blog, we are now preparing to resubmit our proposal to NSF to establish an Industrial/ University Collaborative Research Center jointly with Georgia Institute of Technology and the University of California, San Diego with a hopefully more positive response since our partners now have obtained the minimum number of letters of interest from industry, something they were unable to deliver the last go around.
Planning has started for a Multicore Applications workshop/conference to be hosted by MC2 in late Aug./early Sept. and we will let all know as soon as plans firm up on the date, speakers and focus of conference. Interestingly, IBM has identified Cell staff experts to work with each of the six projects that we identified as areas we plan to explore with the Cell cluster. Moreover, our Associate Director Prof. Yesha, has been invited to speak before the IBM Academic fellows to provide them with an overview of the MC2. I too have been invited by the Computer Research Associate organization to attend a conference in Silicon Valley by invite only hosted by Yahoo that will focus on Hadoop and Data Intensive Computing, clearly a subject of wide interest among our staff. I hope to bring back some interesting ideas that I will include in my next blog.
To conclude this report, I want to say that while our center is still very young as an organization and our work in this area of multicore computations has started to gain wider recognition, a great deal of effort still lies before us.
Milt Halem, Director