Starting this fall, UT and industry will work together to further innovation in the field of computer systems research through a program called the Systems Research Consortium, launched by the UT Department of Computer Science.
Led by UTCS professors Simon Peter, Emmett Witchel, Vijay Chidambaram and Chris Rossbach, the consortium will bring companies that rely on computer science together with researchers on campus to help guide research at UT and introduce problems that need to be solved in the industry. So far, Microsoft, Google and telecommunications company Huawei have joined the consortium as industry sponsors.
These types of partnerships not only allow academic researchers to work on interesting problems, but they give opportunities for industry to benefit from cutting-edge research and student talent.
“The systems research consortium is a win-win,” says Peter. “It’s a win for industry because they will stay abreast of the latest research that we are doing, and at the same time we are staying up to date with the real problems that industry sees, which is important for us because it helps us steer our research a little bit.”
In software engineering, systems refers to the organization of operating systems, which support all the other software applications on a computer. Systems research focuses on developing ways to make software run more efficiently through the organization of these operating systems. This research can be used to improve applications such as data centers and communication networks.
“Systems is extremely relevant to industry from a very practical standpoint,” says Rossbach. “No matter where you’re operating in industry, if you’re somebody who builds web applications or runs a service, you care about a whole bunch of things underneath that which are the subject of systems research. Research in that area will make it easier for you to build things that function in a reliable way and are efficient.”
Chidambaram says many common online services, such as Google and Facebook, have extremely sophisticated systems, and research is vital to improve these services.
“Our world is exploding with information,” Chidambaram says. “Handling this much data and making sense of it requires robust systems infrastructure with high performance.”
According to Peter, systems research is highly practical and has immediate relevance in the real world. The consortium will allow researchers to hear about problems and trends in industry and anticipate what systems will look like in the future, which is something that most companies do not have the luxury to do.
A major area of systems research is improving the efficiency of data going into and out of servers. According to Peter, this issue is of interest to search engine companies such as Google, because if a search result takes more than a few tens of milliseconds to appear, the user will get bored and use another search engine. However, assembling these search results is very complex and involves contacting hundreds or thousands of servers in a data center.
“Each of these lookups that has to happen will contribute to the overall latency to serve you this page,” explains Peter. “This means for each of these hundreds of backend servers, the times that they have to be able to react in has to be even smaller, on the order of microseconds. We’re working on trying to cut down on these service times for these hundreds of backend servers … so that the web page can be presented to you, the end user, on the order of milliseconds.”
The “cornerstone” of the consortium is an annual workshop series hosted at UT, the Texas Systems Research Symposium, starting in the spring of 2018. The symposium will be an opportunity for industry partners to discuss technical challenges they encounter and for faculty and graduate students to present their research.
“It’s very important to us because it helps us to stay on top of the problems and trends that industry sees,” Peter says. “It’s also very important to industry because that’s the main means of output that we have to tell industry what the latest research is and to have a discussion with them.”
In the consortium, industry partners can send engineers to the university to work together with researchers on-site. The program will also include opportunities for representatives of industry participants to give seminar lectures or guest lectures on campus and recruit students. While the consortium will mostly be for researchers and graduate students, undergraduate students will also benefit from interactions with companies through classroom interactions and events.
“We invite an engineer to come and give a lecture in a classroom or to just come by and talk a little bit about what they’re doing, which can help the students decide if this is an interesting company they’d like to work for,” says Peter. “It also gives them hands-on experience on how the systems concepts that we’re teaching here are being used in practice in industry.”
One factor in the creation of the systems consortium was that UT Computer Science added three new systems faculty - Peter, Rossbach and Chidambaram - in 2016. Peter, whose PhD advisor at the Swiss Federal Institute of Technology in Zurich had previously worked in industry and started a similar program there, says he realized the importance of collaboration with industry and knew he wanted to do something similar at UT.
“Our students do internships at these companies, they take the technology that we develop in the lab, they bring them into the company, they can potentially bring back interesting problems and new challenges,” Witchel says. “That’s where the magic happens in these relationships.”