Partners in networking speed

Transcript

RACHEL ROBERTSON: This is Rachel Robertson

JOHANNA CARSON: And this is Johanna Carson

[MUSIC: "The Ether Bunny," Eyes Closed Audio, used with permissions of a Creative Commons Attribution License]

NARRATOR: From the College of Engineering at Oregon State University, this is "Engineering Out Loud"

ROBERTSON: Uh, is that your phone?

CARSON: Oh yes, sorry about that.

ROBERTSON: So, what kind of phone do you have there?

CARSON: That is an iPhone 7 plus.

ROBERTSON: And why do you need such a ginormous phone?

CARSON: I need this tablet, phone, TV because I love photography

ROBERTSON: Okay, so what do you do with your phone?

CARSON: Ah, well. I take photos for myself, for work, and then I post them on Instagram, Facebook, other apps. I even have other apps that I use to process the images, much like you would used to use, perhaps, Photoshop. Now I do it all of that on the phone quickly and I upload them to whatever social media I like to use.

ROBERTSON: And so when you using all those apps, does it always go smoothly or are there times when it’s not so great?

CARSON: It goes smoothly if I’m at home or I’m at work but, for example, if I’m traveling maybe I won’t have as much of a consistent internet access so it might not work as smoothly. And what is interesting to me is how quickly I can become frustrated if I don’t have that immediate access.

ROBERTSON: Exactly, and our obsession with social media and video streaming have already put stress on our networks and could create some big communication problems in the future.

So Cisco, a network manufacturing company, estimates by 2021 the amount of data traveling through the internet will be a billion trillion bytes.

CARSON: A billion trillion.

ROBERTSON: I don’t even know what that is.

CARSON: It’s hard to fathom.

ROBERTSON: I feel like this is something we don’t really think about. We expect things to work even as we put more and more demand on our communication systems. It’s when things don’t work like we expect that we realize that there are limitations to the system.

[MUSIC: Getaway Feat, Ethereal Delusions and Flammen, used with permission of the artist.]

But there are researchers here at Oregon State who are thinking ahead for us and are working overcome the challenges of the future. A lot of this this research is enabled by a partnership with Tektronix that I’ll explain more about later in the podcast.

First I want you to here from with three researchers from the College of Engineering. I asked them what the future of networking will look like. Matt Johnston, assistant professor of electrical and computer engineering, gives us his vision of the future.

MATT JOHNSTON: The idea of where data is stored will be something you don't think about or something that is somewhat meaningless.

ARUN NATARAJAN: Yeah, that’s right.

JOHNSTON: So even now, you know 4G wireless is very fast and so I can stream music I can stream video and that works seamlessly. But I still know that I'm streaming and I still know how much space is left on my phone if I want to store music there. And whether it's on your phone or whether it's on your computer, I think that once wireless data has the bandwidth, and there's also the ability to move all the data around in the back end, you won't see this anymore. You won't know where your data is stored, and you won't really care where your data is stored because it will always be accessible. And it will be all of it -- all of your photos, all of your files for all time.

ROBERTSON: So 5G

NATARAJAN: Sure.

ROBERTSON: What what is that going to bring for us? Or what complications does that bring for you?

JOHNSTON: Is it just faster 4G?

NATARAJAN: They would tell you that it's not. Initially it's going to look a lot like that but there are some fundamental differences to the way that we do things in networks now.

ROBERTSON: That was Arun Natarajan, an associate professor of electrical and computer engineering who is going to step back and give us a little history.

NATARAJAN: Earlier networks, I mean, I'm talking maybe 50 years ago, our wireless network was essentially a TV station broadcasting a program and that was the wireless network that we had -- the communication problem that had to be solved. And then we moved on to the world of cell phones and I should note that sometimes we tend to take the advances that a cellular network has brought to this world for granted. It's actually quite amazing if one really think — I grew up dialing a rotary phone. But now when I go back to India I see people carrying two, three cell phones at a time. And it is just completely revolutionized — revolutionize the way that we do applications the way that we run businesses the way that we personally communicate. And what 5G is trying to look at is this next world where it's not just people talking to people it's machines talking to machines.

[MUSIC: Getaway Feat, Ethereal Delusions and Flammen, used with permission of the artist.]

ROBERTSON: What Arun is talking about here is the Internet of Things. If you are not familiar, the Internet of Things is a general term for all the internet connected sensors and devices that are constantly beaming data up to the cloud. For example, Johanna here … well, Johanna tell us what you can do from your phone.

CARSON: Well, I can unlock or lock my front door, turn off or on any lights in my house or on the outside of my house.

ROBERTSON: So how about your thermostat? Can you control your thermostat?

CARSON: No, not yet, but I’d like to. We’re working on that.

ROBERTSON: Now think of the state of the internet when everyone catches up to technophiles like you.

CARSON: That’s a lot of beaming data.

NATARAJAN: All these devices are entering our network. So how do we accommodate them? And anyone who has been in the airport and tried to access the WiFi knows that it's not easy to accommodate multiple users at the same time. In a stadium our data rate goes down. When there are many users so how do we accommodate all these devices that are entering our network? How do we secure them? There shouldn't be any latency in the system. I want to be able to get within a millisecond the information that I wanted. And then there's also fundamentally, you know, we do need a faster 4G because right now 4G in an open environment is maybe megabits to tens of megabits per second but what we want in 5G is to go 100 megabit per second, a gigabit per second.

So these are the problems that 5G is trying to address and the important thing is it's trying to address all of them at the same time. You want to have more users, more machines, more things in the network, a faster network, a network with lower delay. All these things while facing this fundamental shortage that we have which is of wireless spectrum. If we look at the physics of wireless transmission it hasn't changed. I mean it obviously never changes.

So over the last hundred years we've used up a lot of the spectrum of what the beachfront spectrum is, which is the frequencies where cell phone operates, where TV operates, those are limited. So how do we fit more users more data into that spectrum those are the problems that we want to solve.

ROBERTSON: As Arun says, the physics won’t change, so what this group of researchers is doing is coming with clever ways make improvements in efficiency to use what we have more effectively. Matt, Arun, and Tejasvi Anand (who you will hear from in a minute) work on different aspects of this chain from data collection to the transmission of the data through both wireless and wired networks. Pushing the boundaries of speed and accommodating more devices, means using more power. So what they focus on is making devices and systems more energy efficient which often means smaller.

Matt works on the sensor end inventing tiny chemical and biological sensors for things like portable medical devices. Arun works on wireless networks.

NATARAJAN: Wireless is magic -- everybody loves magic.

ROBERTSON: I just love that he says that because that’s how I think about it.

And Tejasvi Anand, who is an assistant professor in electrical and computer engineering, is all about wired networks. He will explain why we still need them.

TEJASVI ANAND: So basically my research sits on the other side of the spectrum of what Professor Natarajan does and he's more into wireless communication, communicating data from point A to Point B but I'm more into the wireline communication. And the reason is that the high data rates it's very very challenging to communicate from point A to Point B wirelessly in a very energy efficient fashion. And what wireline gives us it gives us a dedicated medium. For example, air is a medium for wireless communication and it is shared. But if you have a dedicated wire which goes from point A to Point B then you can use that medium whichever way you want and we use it completely and make our systems more energy efficient.

JOHNSTON: So, are wires and wireless, are you trying to put each other out of business?

NATARAJAN: That's the complete objective.

The problem is that it's like the rabbit that they have in the race to make people run faster. The wired guys just keep getting more and more and more efficient and wireless just has to work really hard to keep trying to catch up.

ROBERTSON: As you can probably tell from listening to them, Matt, Arun and Tejasvi work pretty closely together even though they each have separate areas of research. In fact, they all share a lab with five other faculty members and all of their students which is a total of about 50 people. Plus their friend BERT.

BERT is not who you are thinking of – friend to Ernie. BERT actually stands for Bit Error Rate Tester and it is a sophisticated piece of equipment which tests new electronic devices to see how many errors are made during the transmission of a signal.

BERT is the crown-jewel of recent upgrades to the lab that were made possible by a host of people across campus and at Tektronix. The reason Tektronix is an excellent partner is because they design and manufacture state-of-art equipment for testing and measurement that essential for research in electrical and computer engineering. They’ve been at it a long time too … 70 years.

ANAND: So, BERT is a kind of equipment that can transmit our data. The one that Tektronix gave us, it can provide the data at the speed of 28 Gbps (gigabits per second) and it can also receive the data at 28 Gbps which is generated by our transmitter, the chip that we design. So, it basically helps to characterize the systems that we design to see how good our system is. So just to give you a perspective one of the key metrics for a wireline communication system is how many errors you have done while doing this communication. And most of the time just because of the high data rates we can tolerate an error if an error happens once in 10 to the power of 12 bits. So that is one bit in error in every thousand billion bits or one bit in error in a trillion bits. So that that's the kind of accuracy the wireline communication should have.

ROBERTSON: The key feature of the BERT is it’s ability to test at a high data rates. Although right now they are testing at 28 gigabits per second, they will also be able to test at 56 gigabits per second which is the upcoming standard. Arun explained why the BERT and the other pieces of equipment were essential.

NATARAJAN: We can't do good research without good equipment. It’s just not possible. I mean ultimately the business that we have got to validate our ideas an there is no validating them without having state of the art equipment and having Tektronix as somebody that's been pretty generous towards our lab in the past has really helped because it helps us demonstrate some of our ideas in practice which is what gives the idea credibility in our space. And the equipment that we've gotten it's been across a spectrum of applications from the bit error rate testers to oscilloscopes to arbitrary form generators which are all equipment that I need in my research for example and it is helped us, again, validate and publish our ideas.

JOHNSTON: And I think what's great is if you're going to like a company like Tektronix who really produces state-of-the-art equipment across all of sort of test and measurement. And so that's why the three of us who work on relatively disparate ends of the spectrum, literally, all got real benefit from this. So far from 28 gigabits per second, you know, we worked around it at DC, Direct Current. So if you think about it like a household outlet if you're running a toaster and that's something like 10 amps, 10 amperes, right, and if you're looking at a laptop it's a couple of amperes. For a lot the sensors we're building because we're trying to get very small and very energy efficient. We're looking at nanoamps which is a billionth of an amp. And it turns out that measuring a billionth of an amp accurately and reliably is a very, very difficult thing to do and requires a very advanced state-of-the-art equipment. And so that also came as part of this part of the package of equipment.

[MUSIC: Getaway Feat, Ethereal Delusions and Flammen, used with permission of the artist.]

ROBERTSON: So, I think we could nerd out about equipment all day, but the interesting part of this to me is what they can do with it. They are all prolific researchers, so I can by no means give you a comprehensive look at what they do, but I wanted you to hear from each of them about how they are working to make devices and systems more energy efficient. We’ll start with Matt at the sensor end of the spectrum who will get into the details of how he reduces the amount of power that sensors need to operate.

JOHNSTON: So from a data perspective the way that we decrease energy consumption and move towards more efficient systems, one of them is being a little smarter about how you decide what to transmit. So it takes some amount of energy to run a sensor in to get data. It takes some amount of energy to compute that data locally and get some information from that raw data and it takes some amount of energy to transmit that to your nearest base station, if you're if you're working over a wireless link. Now it turns out that in a lot of sensor systems it uses a lot of power to transmit over a distance. And so you can be smarter about how you compute the data locally which takes energy but less energy so that you can be looking for, say, specific events.

So one of the things we work on is radiation sensors. Now, you could imagine constantly transmitting how much background radiation you're measuring but that's very power inefficient because that number doesn't change very often. And each of those transmissions is costly from an energy perspective. Now if what you're looking for is an anomaly, right, a spike in radiation and that's all that really matters. Then you can sit there and the sensor can locally gather this data. Look at this data and only when it sees a spike or something that the external world may need to know about then it transmits. You know, it's an alarm system. And so, even with that sort of simple change you can really increase the battery power or sort of the battery life of a local sensor.

Another way that we focus on energy efficiency in the context of biological and chemical sensing is purely in size reduction. So if you think of sort of the classic high school chemistry classroom everything is done in a in a beaker, right, in a large volume. And you have to heat these large volumes you have to transport these large volumes and measure these large volumes and all of that takes actually quite a bit of energy. In a modern biology laboratory you'll see this is all done in milliliter scale — in very small, small volumes. But some of the platforms we work on are trying to recreate these experiments. But at nanoliter volumes so a billionth of a liter and in that case now you need to heat it or move it around. That takes way less energy than even a modern milliliter scale. So that's both of those both in just reducing the size of the things we need to heat and cool and move and measure as well as being smarter about how we handle that data both locally and how we transmit that data wirelessly.

ROBERTSON: Being smarter with selective listening is one research focus for Arun also, but applied to wireless networks.

NATARAJAN: As I was describing the network that is getting more congested. What that translates to is many devices are all transmitting over each other, talking over each other essentially. Is that I have to talk louder and louder to get heard myself. And that takes up energy on the on my device on my cell phone. The other way to look at it also is that if everybody is yelling and my desired signal is also very small I need to work much harder to be able to just listen to what I want to listen while rejecting everybody else. So that takes energy as well.

From the integrated circuit perspective what we're trying to do is build better what we call filters — essentially things that reject undesired signals and pick only the desired signal. So we focus a lot on building integrated filters to overcome this problem.

The other end of the approach is that OK so I need a certain amount of energy to transmit if I use batteries. My sensor dies after some time. How do I keep up a perpetually powered sensor? How do I basically provide them with infinite lifetimes. And the way that we do that is to either energy harvesting or wireless powering.

ROBERTSON: And once again, smarter techniques are part of saving energy for wired networks. Here’s Tejasvi.

ANAND: Even when there was no useful data that is being transmitted, the processor transmits some junk data to the memory just to keep the communication channel open and that consumes the same amount of power.

And so scientists at Google they figured out that most of the time these links are used let's say 10 to 15 percent of the time and they were idle for 75 to 80 percent of the time this communication channel was not being used.

So in one of the research direction we try to reduce that amount that power consumption by just shutting off of the wireline links when there is no data to be transmitted. And then and then bringing it up instantaneously whenever there is a data that is required to be transmitted. Now it may seem very trivial that they're not doing it in the beginning. They should be doing it like this. The biggest challenge in this is that all this turn on and off has to be done in nanoseconds. Or one in a billion seconds. And it is it is of that order that makes this entire turning on and turning off very challenging and we basically were successful in doing that. We were able to turn off and turn on the link in less than 20 nanoseconds. And by doing that we were able to save a lot of power.

ROBERTSON: Tejasvi did some calculations to find out what kind of impact the energy saving techniques he has developed would have. He found that by removing idle power, the data centers in the U.S. could save 1.4 gigawatts of power per year which would translate into $800 million dollars. So to give you some perspective, one gigawatt of power could charge roughly 250 million cell phones.

So, you can see where companies like Facebook and Google, which have large data centers, might be interested in the technologies that Tejasvi is developing. But it might be less apparent why Tektronix, who manufactures testing and measurement equipment would to partner with us. To find out why this is a good collaboration for them as well, I went to Tektronix in Beaverton to talk to Brandon Greenley

BRANDON GREENLEY: I'm the vice president and general manager for our components business at Tektronix.

ROBERTSON: But more than that. Brandon is a proud Beaver, who received both his bachelor’s and master’s degrees in electrical engineering here and has remained involved with Oregon State as a part of university recruiting for Tektronix.

GREENLEY: We want to continue to support and see Oregon State be the engineering pioneer that it is in the state. It continues to be our go-to school for a lot of different reasons. So we want it to be strong, and continuing to support it, from an equipment standpoint means we move research forward and we just raise the bar for the entire department. So, that's our main purpose is to be feeding back into the community, bettering that community through a lot of different actions. This is one of those.

ROBERTSON: In fact, Tektronix has had a long established partnership with Oregon State. They provided another equipment upgrade about 10 years ago and have supported innovative education and student engagement. In fact, Brandon was part of Tektronix involvement when he was a student. His internships with them led to his later employment at Tektronix. So, he is a pretty big fan of partnerships between industry and universities.

GREENLEY: I hope that Oregon State can continue its leadership in the university community at fostering these industrial partnerships. I don't see it at all the other schools around the country. Now that's maybe to the benefit of Oregon State, but I hope that we don't lose that recipe.

ROBERTSON: If you ask Arun, it is a recipe worth keeping.

NATARAJAN: I think our research in general is a very practical area of research, a very applied area of research where we're trying to build real systems and demonstrate capabilities that impact real applications. And in order to do that we have to be close to industry at some level because that's the way that our technology moves to them and their technology comes to us. And I am cognizant of the fact that the primary thing that really comes out of our lab is our students. That's what we are set up to do. And in order for them to get the best out of their research both now and in the future, this kind of close interaction is required and which is something that the integrated electronic area has consistently had. If we look at our support for our programs it is been very industry heavy which basically at some level tells us that what we're doing is relevant.

ROBERTSON: That was a great summary of why partnerships with industry are essential to Oregon State for both research and student engagement with future employers.

I should mention that there were a lot of people that were involved with the upgrades to the Analog/Mixed Signal Lab including Tom Weiss at Tektronix, and Julie Brandis and Un-Ku Moon at Oregon State who all worked together to make it happen. Donations and discounts were provided by Tektronix, and additional funds were provided by the researchers, the College of Engineering, the School of Electrical Engineering and Computer Science, and the Research Office at Oregon State.

Thanks for listening, my friends. I hope you enjoyed hearing more about how researchers at Oregon State are out there changing the future.

[MUSIC: Getaway Feat, Ethereal Delusions and Flammen, used with permission of the artist.]

This episode was produced by me, Rachel Robertson, with lots of help from my friends. Audio editing magic was performed by Brian Blythe. Our intro music is The Ether Bunny by Eyes Closed Audio on SoundCloud and used with permission of a Creative Commons attribution license. The other music in this episode was Getaway Feat by Ethereal Delusions used by permission of the artist. For more episodes visit engineeringoutloud.oregonstate.edu or subscribe by searching “Engineering Out Loud” on your favorite podcast app.

Find out more about Ethereal Delusions by following Twitter @quixoticBPMoose.