A student at Duke University in North Carolina has developed a modification which he claims can boost the battery life of mobile devices with just a simple tweak to the Wi-Fi driver. We chat with Justin Manweiler about SleepWell and the problems he's trying to solve.
A fifth-year PhD candidate in Computer Science, Manweiler has long been interested in wireless technologies. "I've been studying wireless networks for about four years now - it's my whole thesis area," he told thinq_ during our interview. "We've sort of figured out performance in a lot of ways, we're giving users the capacity they want - at least from Wi-Fi. I won't speak about cellular in general, but in terms of Wi-Fi people are getting what they need out of it."
With performance a more or less solved problem, Manweiler looked to other areas where he could innovate. "What is the real-world problem that users are suffering in mobile today? I really think that's energy right now. The real problem is that people's phones are just dying, the battery life is just dying too fast."
That's something with which any smartphone owner can agree, but is surprisingly something that wireless manufacturers have been lax in addressing. "The most basic power saving mechanism in 802.11 - or Wi-Fi as its more commonly known - is something called, simply enough, 'Power Save Mode,' or PSM," Manweiler explained. "What PSM allows the clients to do is to listen to these periodic messages from a router or an access point that announce 'I have traffic for you' or 'I don't have traffic for you.'
"What happens is the clients actually go into this really low power sleep state when they don't have traffic, and then once in a while when the access point does say 'I have traffic for this client,' the client stays awake and downloads all that traffic."
It's a neat solution in theory, but one that Manweiler claims doesn't work all that well in practice. "The problem is that a lot of these techniques that were developed weren't really designed in view of the way networks really operate," he claimed. "What I mean by that is the kind of device and infrastructure density that we really see right now - lots of smartphones, lots of laptops, and this sort of thing.
"So, we have a lot of these different devices, a lot of access points, a lot of routers, all trying to use the same wireless channel - and that channel is a scarce resource. What happens is that as we have traffic congestion, as this traffic kind of contends for the shared wireless channel, the power saving optimisations break down."
That's the problem that Manweiler's creation, SleepWell, looks to solve. "The project is all about whether we can reimagine how these power saving optimisations should work, if they had been designed in light of realistic networks and in light of device density and the kind of heavy utilisation that is truly commonplace today," he enthused.
"The really cool idea behind SleepWell," Manweiler explained, "is to say 'can we be a little bit smarter about how access points advertise available traffic' - when should they advertise that traffic, when should these beacon messages be transmitted' and a few other things on top of that. But it's really a minor tweak on top of the way that PSM works which adds a little bit of scheduling that cuts down on the cost of this traffic congestion that we see."
The elegance of SleepWell is in its simplicity. Rather than requiring new hardware or complex coding, Manweiler's technology can be implemented with a simple change in the driver software to allow the access point to 'lie' about the time in order to reschedule the wake-up signals. "My implementation - I have a complete and fully-working prototype - is simply written into driver software," he claimed. "I didn't need to change firmware, even. It's a pretty small amount of code - I mean, the ideas are pretty simple.
"That said, they're effective, and really I think that's the ideal: small changes that anyone deploying an access point or anyone who's selling an access point can really implement very quickly and get the benefits, without breaking standards compatibility in any way or having to change hardware or make any major changes to the entire software stack on their access point."
A router or access point equipped with SleepWell technology would, Manweiler claims, be 'battery-life-aware,' boosting the run time of portable devices connected to it. The technology isn't a panacea for the growing power draw of modern 'superphones,' however. "There are some cases in which it doesn't give you all the benefits and adds a little complication," he admitted. "There are certain types of traffic that don't work very well with Wi-Fi Power Save Mode - things that are highly latency sensitive like Skype or where you have really interactive traffic.
"That's not something that SleepWell can really help too much. We're talking about scheduling that's more coarse-grained than that," he explained. "There are other optimisations that might be appropriate there, some of the things that are already in the existing Wi-Fi standard but not yet widely implemented will help."
Instead, SleepWell targets latency-insensitive but high-bandwidth applications. "The targeted applications are things like heavy-bandwidth apps like YouTube, file transfer, Internet radio, and a variety of the other types of less latency sensitive applications that we run on our mobile phones constantly." Even with this limited subset of possible applications, SleepWell promises major savings.
"I had an experiment where I was downloading a YouTube video on two phones simultaneously - a couple of users, and very normal application, YouTube, and we still see substantial energy gains," Manweiler claimed. "Y'know, if you're doing something even more strenuous on the network - let's say downloading a large file - the gains can become quite large from just a little bit of traffic."
Those gains become sizeable as the congestion on the network increases, allowing client devices sleep through traffic bursts meant for other users and for other access points. "I won't say this is only suitable for a congested environment where you have lots and lots of devices," Manweiler explained, "but the probability is that it will be a help - and the amount that it will help will certainly increase with density.
"In cases where you're the only device around - y'know, the PSM method as it was originally envisioned will be fine, so you're not losing anything. We're not throwing any overhead into the system, it simply makes things work a little better when we do have these congestion problems."
While the project is gaining interest - Manweiler's research on SleepWell was nominated for the Best Paper award at the MobiSys conference, and he scored a runner-up place in the Best Demo category - it's a long way from becoming part of the Wi-Fi standard. "We're really only at the beginning steps of this," Manweiler admitted. "I mean, the first step is really just to disclose this to the research community, which happened last week at the MobiSys Conference. At this point we're thinking about, y'know, how do we get impact from this, how do we get this out in the field and benefit users? That's really the stage that we're at now."
Although Manweiler has approached the Wi-Fi Alliance, the industry body which sanctions modifications to the 802.11 standards, he's hoping that manufacturers will be able to implement SleepWell without the need to ask permission. "Our first plan is to contact the vendors, because it's really the wireless vendors that sell the equipment that really stand to deploy the technology en masse.
"The nice thing, if you think about it from a commercial perspective, is that if you sell access points and you deploy an entire building, you can get all the gains that it offers inside that building without having to talk to any other manufacturers or any standards bodies. It would be a major selling point: selling the first access point that is battery-life-aware for the clients attached to it and associated to it. I mean, I think that's an upgrade reason in and of itself."
Manweiler's paper on SleepWell can be found over at his Duke University homepage.