If you’ve ever used a large-scale web application like Amazon.com, you might have noticed that sometimes certain pages open quicker than others. There’s a reason that happens, and one of the people trying to solve the problem is Qingyang Wang, assistant professor of Computer Science at LSU.
Wang’s research focus is cloud computing, and this semester he is teaching a Cloud Fundamentals and Web Programming class. His current research project could help companies give better guarantees about the performance of different applications in the cloud. “Cloud computing is the current and future computing platform for all the important enterprise applications,” he says. “So, it’s an important specialty for students in the current and future job market.”
We recently spoke to Wang about his current research in cloud computing.
Seeking Ways to Conserve Energy
As cloud computing grows, there needs to be a more efficient way to power it and still provide strong performance for users.
Enterprise cloud data centers — the computing infrastructure built by large companies like Amazon, Google and Microsoft — have thousands or even millions of physical machines to host large websites. However, those machines only run at an average of 15 percent of their capacity so that end users have less trouble with pages opening slower. For an e-commerce site like Amazon, slow internet response time for users leads to a drop in sales. However, this strategy wastes a lot of computing resources and energy.
The variation in web response time that users experience is called long tail latency, and it occurs when the machines powering the cloud reach higher levels of utilization. To avoid the long tail latency problem, cloud data centers keep their machines utilization levels around 15 percent, which means 85 percent of the computing power is wasted, Wang says.
Research Project Focus: Solve the Long Tail Latency Problem
Wang has been working on a project funded by the National Science Foundation, which seeks to solve the long tail latency problem of web applications in the cloud in a more efficient way.
His research has two areas of focus. First, to investigate system-level factors that cause the problem, and second, to design and implement solutions that will achieve a fast response time for users, even when a high number of people are using the site. The hopeful outcome will allow the machines supporting cloud data to run at a higher utilization level and prevent so much wasteful and costly energy consumption, he says.
He’s been working on this project for about a year and expects it to run another year or two depending on progress. There are currently two PhD students assisting him, and undergraduates help on a semester basis.
The Future of Cloud Computing
Using applications and storing data in the cloud can be a cost saver for small to mid-size businesses.
Someday, cloud computing will probably be the way all computer work is done, minimizing PCs to just a screen and keyboard.