Skip to main content

CableLabs Gives Latency a Licking

The race is on to deliver the Internet at speeds of 1 Gigabit per second or more, but CableLabs is working on features and techniques for cable’s DOCSIS platform that are designed to help operators boost performance in a way that could differentiate their broadband services from those delivered over all-fiber networks.

One exploration well underway is the implementation of Active Queue Management (AQM), a technology that’s designed to reduce latency, buffering and packet loss, therefore improving the overall performance of DOCSIS-delivered broadband services. While raw speed remains key for big media downloads, the reduction of latency can boost the performance of a variety of Web-delivered applications, including multiplayer gaming, video conferencing, video streaming and even the simple task of loading Web pages.

AQM keeps close tabs on how rapidly the buffers on cable modems and cable modem termination systems (CMTSs) are filling and drop just enough packets to ensure that appropriate buffer levels are maintained, according to CableLabs. The tech consortium estimates that proper use of AQM can reduce buffering latencies by hundreds or thousands of milliseconds, which can quickly add up and result in faster load times and a reduction in delays that can be associated with interactive Web apps.

The implementation of AQM “could have a dramatic impact on the user experience,” Dan Rice, CableLabs’ senior vice president of network technologies, said.

AQM has been around for more than a decade, but Cable- Labs has been ramping up its research recently to determine how it could be optimized for DOCSIS links, including those piped in through DOCSIS 3.0, the current platform that’s capable of delivering downstream bursts of more than 1 Gigabit per second, as well as DOCSIS 3.1, an emerging spec that is targeting multi-Gigabit speeds.

Looking to deliver lower packet loss and lower latency, CableLabs has amended DOCSIS 3.0 with a recommendation that AQM be added to existing gear, via a firmware upgrade if possible, and has mandated AQM in DOCSIS 3.1 equipment — both the cable modem and the CMTS.

On the cable-modem side, CableLabs has pinpointed support for an algorithm developed by Cisco Systems called Proportional Integral Enhanced Active Queue Management, also known as “PIE.”

For the CMTS, CableLabs has left it open to vendors to select the AQM algorithm that best fits their architecture.

“The bigger picture is that all across the Internet, every network element should be implementing AQM,” Greg White, CableLabs’ principal architect, said, noting that routers and switches can use one of several flavors of AQM and still provide the intended results when working with cable modems that use PIE — a higher-quality broadband experience.

Operators want to add AQM to deployed equipment as soon as possible, White said, and vendors believe it’s technically feasible to add PIE to 3.0 modems.

“We have test plans written,” White said. “It would be a fairly quick process, and we feel that cable is taking a lead here in getting these current generation AQM algorithms out there. As far as I know, DOCSIS is the only [specification] that’s made a decision that this [AQM] is critical.”

And it’s increasingly important as cable aims to deliver faster speeds as MSOs continue to face off with existing fiber-based competition from telcos such as Verizon Communications, and the emerging threat of Google Fiber.

“You need big buffers to deliver 1-Gig,” White pointed out. “At some point, the basis of competition could shift from speed to other attributes such as responsiveness and performance.”

And consumers are growing more attuned to how well their broadband services perform beyond raw speed numbers. “The State of the User Experience,” a new study from contentdelivery network specialist Limelight Networks that surveyed 1,115 consumers, found that most survey respondents (52%) view fast load times and a buffer-free experience as key to overall performance, while 60% indicated that they aren’t willing to wait more than five seconds for a webpage to load before they would leave the site.