A new article from MIT News recently drew attention to what could end up being big news for the data center services world. A new MIT system, called Fastpass, could reduce data-transmission delays across server farms by 99.6 percent (yes, you read that correctly).
Currently, data centers are prone to congestion. In MIT News’ words, “Packets of data arriving at the same router at the same time are put in a queue, and if the queues get too long, packets can be delayed.” But at ACM SIGCOMM 2014, the flagship annual conference of the Special Interest Group on Data Communication, a vital special interest group of the Association for Computing Machinery, MIT researchers plan to announce a new network-management system that could all but do away with that aforementioned congestion.
Fastpass “relies on a central server called an ‘arbiter’ to decide which nodes in the network may send data to which others during which periods of time.” In experiments, Fastpass:
These results are impressive—even if the system is somewhat counterintuitive. Hari Balakrishnan, who coauthored the paper on Fastpass, remarked that “It’s not obvious that this is a good idea” upon first glance. Balakrishan continued on to note that “This paper is not intended to show that you can build this in the world’s largest data centers today. But the question as to whether a more scalable centralized system can be built, we think the answer is yes.”
Obviously, there’s still a ways to go before we see widespread adoption of this network-management technology. That said, eliminating latency seems to be the new frontier for data center services providers and researchers across the world, and this technology has the potential to push that frontier much farther than previously thought possible.
The era of the no-wait data center may soon be upon us. We look forward to the research team’s presentation, and can’t wait to see what a queueless future may look like.
For more information on Fastpass, we recommend giving the full paper, accessible on MIT’s Open Access Articles database, a read.