You are here
Got Bufferbloat? Prof. Vishal Misra’s Work Is Behind a New Fix! (Columbia Engineering)
Source: Columbia Engineering
There are few things more frustrating than being on an online real-time phone call when it suddenly gets the jitters, your caller’s voice doesn’t match up to her facial expressions, and your conversation doesn’t sync correctly. Gamers are infuriated when sudden lags cause them to lose at what they’ve been playing for hours. These lags and jitters are caused by a phenomenon known as “bufferbloat.”
Bufferbloat is an issue in packet-switched networks, in which excess buffering of packets—meant to keep a congested link as busy as possible—causes high latency and packet delay variation, or jitter, while also reducing the overall network throughput. When a router device is configured to use excessively large buffers, even very high-speed networks can become practically unusable for many interactive applications like voice calls, chat, and even web surfing. And this can occur when there is more than enough bandwidth available to support these applications: in order to achieve low packet loss, the buffers become “bloated” and introduce unnecessary latency.
“This is a problem with modern Internet routers, including cable modems, which all have large buffers built in to absorb the variations in the traffic and congestion levels on the Internet and prevent packet loss,” says Vishal Misra, associate professor of computer science.
And now there is a fix for bufferbloat based on research done by Misra when he was a PhD candidate at UMass-Amherst. The latest version of DOCSIS (Data Over Cable Service Interface Specification)—the software that runs on all cable modems—involves the implementation on routers of Active Queue Management (AQM), an algorithm devised by Misra and his colleagues that enables users to simultaneously achieve high throughput, low latency, and low packet loss on Internet links. The AQM that has been selected for the new DOCSIS 3.1 deployment is called PIE (Proportional Integral Enhanced) and has been developed by Cisco. DOCSIS 3.1 is now being deployed in all cable modems around the world, and is expected to be in more than 100 million units that will be shipped later this year, across all manufacturers.
“The PI (Proportional Integral) controller is what developed from my PhD work, and Cisco’s PIE controller grew out of this research,” says Misra. “There’s an effort going on now to make Cisco’s PIE the default AQM on all Internet routers, not just cable modems, so what’s very exciting is that pretty soon a wacky idea I had in class one day might impact all traffic on the Internet around the world!”
The foundational work for Cisco’s PIE was done as part of Misra’s PhD thesis work on TCP modeling. Transmission Control Protocol is the protocol that carries nearly all of the traffic on the Internet and is key to understanding and solving the problem of bufferbloat. Researchers at Cisco have been working on a bufferbloat solution for several years and looked at many different algorithms.
As a PhD student at UMass-Amherst, Misra was listening to a lecture on stochastic differential equations by his PhD advisor, Professor Weibo Gong, when he realized he could use similar equations to model TCP on the Internet. So he approached Professor Don Towsley, his co-advisor and an expert in network modeling. Towsley was intrigued as he had recently written a paper on modeling TCP throughput and had a lot of detailed TCP measurements. “I gave the data to Vishal to validate his model,” Towsley says, “and he came back with a careful statistical study that showed a great match.”
At the SIGCOMM conference in 2000, Misra, Gong, and Towsley presented a highly influential paper on the TCP model that has since received over 1500 citations and forms the basis of TCP analysis by almost every researcher around the world.
“What separated Vishal’s model from what had been done prior to that,” notes Gong, “was the fact that it opened up entirely new directions in TCP analysis, from understanding its dynamics to developing fast simulators.”
With the publication of his paper, Misra realized that the way congestion was controlled on the Internet was inefficient and his model could help develop better control mechanisms. He decided to approach Professor C.V. (Kris) Hollot, a control theorist at UMass with whom he had not yet worked. Intrigued by Misra’s model and its potential for formal control theoretic analysis, Hollot attended Misra’s PhD defense and was drawn in immediately.
“Vishal and I often joke that I was his first graduate student,” Hollot says with a laugh. He, Misra, Gong, and Towsley went on to produce a series of papers that have together received more than 3000 citations. One of the first papers they wrote was on the Proportional Integral (PI) controller. In this study, they demonstrated that by observing the direction in which the router buffer queues change, they could provide a feedback signal that could proactively prevent congestion. They also ensured that their controller algorithm was inexpensive to implement on routers.
But nothing much happened in the commercial world until 10 years later, in 2010. By this point cheaper memory and faster links had made the problem of bufferbloat even more acute, and CISCO researchers started taking a closer look at the work by Misra and his colleagues. In 2013, their new system, which they named PIE, was chosen after extensive performance testing over other competing mechanisms to emerge as the AQM of choice for cable modems; it has now been made the standard for the next version of cable modem software.
“There is another brilliant effort called CoDel, designed by some of the pioneers of the Internet, that matched PIE’s performance,” Misra says, “but PIE was so simple to implement that it was the top choice, thus validating the design we developed more than a decade ago.”
“It is very gratifying when you learn that your research has made a difference in the real world,” adds Misra. “That’s what you live for as a researcher.”
—by Holly Evarts