Jeremy Craig, 404-413-1357
More and more, businesses and other organizations have made the switch from using their own computing resources, such as servers to software on individual machines, and have moved into cloud computing. With cloud computing, computing resources like data and processing power reside elsewhere, eliminating the need for so much hardware on site – helping the bottom line.
Now, Georgia State University researchers and staff are working to make the clouds more efficient and less prone to errors, helping to save time and money.
In cloud computing, customers purchase processing time, memory or disk space for use in various applications. And there’s a need to detect which parts of the cloud have gone bad – something called “fault detection.”
GSU researchers are using control theory – which works by taking constant inputs from one source to make needed changes, just like how a thermostat controls air conditioning or heating – to improve fault detection. One way to do this is to test the time it takes for a message to reach one point from another, or measure the response time of a job that the system is performing.
“This way, once you detect one fault, we can isolate, separate and virtualize another part of part of the system for you to use, so it is always smooth and reliable,” said Yi Pan, chair of the Department of Computer Science.
It’s also important to make the use of various resources, such as computers’ central processing units (CPUs), efficiently, which is called load balancing.
“If we have a virtual computing lab [a type of cloud], we have a certain number of CPUS. Some CPUs may be very heavily used when someone is running a certain type of software, but other CPUs might not be used,” said Pan, who recently won an IBM Faculty Award for research in cloud computing.
“We have to detect the load and how to move some of this user’s load to balance the cloud,” he added.
Working with Andy Rindos, a researcher from IBM, Pan said the next step to combine both load balancing and fault detection, making the process seamless.
The effort to bring more efficient cloud computing has come home to GSU, where the Information Systems and Technology (IS&T) division has set cloud computing as a new standard in computing technology at the university. The university purchased 15 virtual servers for administrative computing and also launched a virtual computer lab for students.
The virtual computer lab is a key part of improving student computing options. Instead of students having to be sitting at a computer, in a particular location, with the particular software suite they need, students can request computing set-ups called “images,” each with the particular software they want, right from the comfort of their own homes.
“It’s like having a car, but you can switch out your engine with a different one,” said Art Vandenberg of the IS&T division. “Your personal computer can have a MATLAB engine, a SAS engine or a another suite of software – without actually needing all that software installed directly on your computer.”
Complementing the cloud computing activities of Georgia State University’s researchers and staff, Pan has also launched the International Journal of Cloud Computing, the first such journal in the world devoted to cloud computing.
Published Feb. 14, 2011