Prakashan Korambath, an expert computer programmer, has tackled many challenging projects during his career. For starters, he has helped countless researchers debug their code. He also has helped develop numerous software programs to improve the efficiency of the computation process and he played a major role in UCLA’s transition to a cluster and grid computing service provider.
But however impressive these and his many other accomplishments might be, Korambath believes they pale in comparison to his current project: deploying a cloud storage service for UCLA. The “Virtual Computing Laboratory,” a pilot project of the Instititute for Digital Research and Education, allows users to upload and download data in and out of storage from anywhere a network connection is available.
“Amazon has everything on a public cloud,” said Korambath, a research technologist at IDRE. “We want to do the same thing privately. It’s challenging technology. Storage, security, hacking – if you take the responsibility of storing someone’s file, you have to be responsible. You have to answer the right questions. It takes a lot of knowledge to implement. I’m using all the knowledge background I have so far.”
The cloud storage system uses OpenStack open source software, which allows users to see the code and learn from it.
“Cloud computing is one of the intensive research areas these days,” Korambath said.
A computational chemist who earned his Ph.D. at the University of Memphis, Korambath has contributed code to computational chemistry software programs GAMESS and MOPAC, and was instrumental in parallelizing Q-Chem Quantum Chemistry software. At UCLA, his work has focused on user software as it relates to high-performance computing.
Upon arriving on campus in 2002, Korambath helped university researchers transition from the outdated and expensive IBM Scalable Power Parallel 2 supercomputer to a cluster of desktop computers running an open source Linux-based operating system. For Korambath, this involved compiling scientific code, debugging user code, profiling the code to speed up calculation, and more. For UCLA, this marked the beginning of the grid and cluster computing.
“The first computer cluster we built at UCLA had 18 compute nodes,” Korambath said. “By 2012 this cluster has grown to 750 nodes.”
Korambath has contributed significantly to the evolution of the UC grid system, frequently collaborating with researchers on projects designed to streamline workflow and increase computational efficiency. For example, he collaborated with the Scientific Workflow Automation Technologies Laboratory at the San Diego Supercomputing Center to develop a theoretical enzyme synthesis using the Kepler workflow for researchers in UCLA’s chemistry department. The workflow ran on the Hoffman2 Shared Research Cluster, and was able to process approximately 7 million jobs in a period of 9-10 months. If the process had been done manually, it would have taken about three years.
“Today we have the tools to make it possible to run 100 jobs at the same time,” Korambath said. “We’ve automated the whole process. To get to that point you must run a prototype and make sure what you’re doing is correct. Then you run it again for the entire data. It’s a repetitive operation with many steps involved.”
Since beginning the cloud storage system project, Korambath’s focus has switched from user software to cloud computing and storage. He believes this is the future of high-performance computing. See the Tech Reports section of “Reading Materials” for Korambath’s papers.
“My purpose in writing the reports is so people can read them and comment and help us improve,” he said. “It will be a benchmark. If someone wants to know the best hardware and software for doing the work we are doing, they can read our paper.”