WELCOME TO FLASHMOBCOMPUTING.ORG
Home of the first-ever flash mob supercomputer and your one-stop site for everything you need to know about flash mob computing
May 7, 2004: NEW RELEASE – USF FlashMob RC 4.iso and source code
One night to the flash mob day, a bug fix- USF_FlashMOb_RC_4.iso (230MB) – was released. This update made possible the configuration of individual nodes with different processor types. In addition, this update allows the running of larger flash mobs by power users. The source code for USF_FlashMob_RC_4 is now available.
May 1, 2004: USF Computer Science Masters Programs
Studying with the faculty, staff, and students that spearheaded the next wave in supercomputing would be awesome! Right? M.S. in Internet Engineering offers an extensive plan of study, ranging from combining networking, distributed computing, and internet technology. The new MS program is comparable to the regular MS in Computer Science with more emphasis on internet application development.
April 7, 2004: Flash mob I: The First step toward practical instant supercomputing
Initially, this project began as a dare. How many people can you convince to entrust you with their conventional computers, bring them to a common location to build a supercomputer that can compete with the fastest and most expensive computers in the world? A simple dare amongst a few individuals had now grown into a platform that would enable a large group of people to gather computing powers with the aim of solving a key scientific problem that interests every one of them. Their primary objective was to prove that building a temporary and inexpensive supercomputer that is accessible to the public was indeed possible. The supercomputer would be made up of other ordinary computers and would be developed with the aim of solving a particular problem. Hence, the need for the experiment that was about to go down. Subsequently, the first computing flash mob event was set for the third of April,2004 at the University of San Francisco, dubbed – FlashMob I. The turnout was impressive with over 700 computers coming into play. By 11.00 am all computers were connected to a network switch that had been donated by Foundry Networks. After that, all the collective computer power was pooled by our custom software to work on the Lin Pack Benchmark. The Lin Pack Benchmark is a mathematical application whose main function is to determine supercomputing speed. After a series of tests, only 256 of the computers were able to run the benchmark achieving the highest rate of 180Gflops. However, the computation of speed stopped three-quarter-way because of node failure. The best and complete sequence used 150 computers that resulted in 77Gflops. Regardless of the mishaps, the experiment proved that it is possible to achieve substantial computational speed from a temporary supercomputer. Unforeseen issues and time were the main limitations that prevented the use of all 700 computers. Fortunately, identifying the issues that would hinder the success of this whole experiment was one of the goals of the project. This information would be used to make and improve future FlashMob computers.
The flash mob presented various technical and logistic challenges. For instance, we had to come up with software that could run CDROM disks that had to be inserted into each of the participating computers. This step was crucial, to avoid tampering or losing data on any of the computers. More importantly, we had to steer clear of the hard disc so as to enable all the participating computers to run the same operating system and application at the same time, without having to install anything on the hard disks. This way, it was safe for people to lend us their computers since we only intended to use the memory and processors of their computers. Logistically, due to a large number of volunteer equipment in one location, there was a need for a registration and asset tracking system. This system ensures that everyone gets their stuff at the end of the day. In future, flash mob organizers can also benefit from our software and also use our registration process as a model.
Attendance to the event was satisfactory with people from different walks of life ranging from high school students, college lectures and supercomputing gurus. Everyone in attendance felt like they were participating in a revolutionary event. The event also got the attention of the media, which enabled our project to be shared with the whole world. The response we have received has been very encouraging with a good number of people getting motivated to organize a flash mob computer in their hometowns. The project made people realize that there are many untapped resources in their computers, which are yet to be put to use for a good cause.
Making our software available on our website has enabled scores of people to run their mini flash mobs at their homes, in schools or even at their place of work. A good number of people are eager to turn their ordinary computers into supercomputers. The whole project also created awareness of the potential and the use of interconnecting multiple computers simultaneously to solve a single problem. Some people had never heard of parallel computing until we launched the project. We feel that the project not only introduced a new type of computing but also popularized its existence and made people more aware of its potential.
The first goal of making flash mob computing a reality has been achieved. We have substantiated that making a temporary supercomputer is possible. The next step is even bigger with hopes of creating scientific applications that are built specifically for flash mob computing. There are many easy to run applications for computers everywhere. We aim to create easy to use scientific application for flash mob computing that can be utilized by anyone to perform their experiments. Our vision is for scientists from different genres to adapt their existing supercomputing applications to the flash mob environment. These applications can be created on easy to run CDROMs that can be duplicated to as many copies as the expected computers on the flash mob. If scientists adapt their applications to flash mob, this will create alternative platforms for people who cannot access supercomputers to do their scientific computations. Our website is open to serve as a platform where scientist and individuals interested in building flash mob computers can meet and interact.
The USF team is still planning and reflecting on improvements that can be made to the core flash mob software. The idea is to make the project an “open source” where anyone can download and make modifications to our code. The most significant improvement is to reduce the amount of time needed to setup and configure the flash mob computers so that the computations can be done as fast as possible. In addition, we would like to significantly reduce the chances of failure of the flash mob computers. The benchmark program that was used for the flash mob I would crash if any of the computers on the flash mob failed in the course of the computation. These two improvements will enable the maximization of the computation as opposed to the configuration time, especially because the flash mob supercomputer is temporary.
Flash mob I promoted unity at USF by bringing faculty members and students in the department of computer science together to make the flash mob software. The organizational effort extended over several administrative units and also received great supports from different levels in the university. Both the volunteers and the non-volunteers made it easy for the participants to carry on with their projects.
Above all else, it is of great importance to note that creating a flash mob supercomputer is much about the people who participate in it as it is the technology they offer.
Creators of flash mob computing.
Pat Miller – a research scientist at a national lab and a professor at USF. His Do-It-Yourself supercomputer class turned into a flash mob. Initially, the students were to bring their computers and Xboxes to make an evanescent cluster at each connection. Mr. Pat Miller worked through all aspects of the flash mob software.
Greg Benson – USF associate professor of computer science. He invented the name “flash mob computing.” The wireless flash mob computers were his idea. He worked through the core infrastructure of the flash mob runtime environment.
John Witchel — A graduate student in computer science, spring of 2004. He talked to Greg about the challenges of wireless networking computers. He proposed inviting people that are not part of the USF to join the flash mob to break the Top 500. John used the flashmob I and the flash mob software as his Master’s thesis.