A conjunction of a few new technologies may trigger disruptive changes in the future. This observation was prompted by a talk Steve Perlman gave at the Columbia Engineering School: Benjamin Button, Cloud Everything and Why Shannon's Law Isn't. In it he covers a set of technologies that at first may seem unrelated, but turn out to be deeply related after all, culminating in a realization of the long talked about vision of an application utility, where all applications are hosted and run out of the cloud.
First Perlman talks about the realistic human rendering technology developed at Rearden, his research incubator company. This technology was developed over many years and is the secret behind the wonderful effects found in movies like Benjamin Button. It is now being used in many other films, and promises to revolutionize film making, possibly even replacing actors with computers, in real-time.
The next invention at Rearden is OnLive, a cloud based gaming technology for playing video games over the Internet. This should be impossible, but they use a new video compression technology that makes it possible to play slow twitch games a 1000 miles apart. This takes their rendering technology real-time and interactive. He demo'd complex software running effortlessly on an iPad, but that was really running in the cloud. Their gaming and rendering technology makes it possible to centralize applications and still yield good performance.
The biggest invention in my mind is what Perlman talks about last, which is DIDO, a low-latency high bandwidth wireless network technology that promises each wireless user to use the full data rate of shared spectrum simultaneously with all other users, by eliminating interference between users sharing the same spectrum. A new paper has been released describing this technology: Distributed-Input-Distributed-Output (DIDO) Wireless Technology A New Approach to Multiuser Wireless. It promises submillisecond latencies, with no towers, and no backhaul network.
Independently each of these inventions are cool. A real testament to invention, perseverance, and beauty. But after a while it becomes clear there's a link between all these technologies and that the sum could be way greater than the parts.
The realistic human rendering technology made it possible to create OnLive. Building OnLive taught Rearden how to use the cloud to make a centralized application server infrastructure. The problem was still networking, you can't play fast twitch games or send out live video without a better network. So they created DIDO to solve the networking problem. The result is a true centralized computing model. All your applications could live in the cloud and realistic UIs will be served over a boundless low-latency wireless network bandwidth.
It's a bold vision, but will it happen?
What I think may happen is a bit different. Instead of this technology stack enabling end-to-end centralization and control, thus capturing Apple like returns, what I think it will enable instead is massive distribution. With a ubiquitous low-latency wireless network it becomes possible to bind the massive distributed power of all our compute resources into one unified application fabric. If it works then in essence what we'll have is a giant wired network. What has been missing is a high-bandwidth, low-latency network to make it so all the distributed parts can work together. Maybe that won't be missing for long.
Who knows what the future will bring. It will no doubt be surprising and unpredictable and totally beyond what we can imagine. But these are the kind of technologies the future is built on. Today what we have is the past and the past is not wearing well.
Let's take a look in more detail at the various parts of the vision...
With networking, especially wireless networking lagging all other parts of the great distributed computing puzzle (CPU, RAM, storage), this new wireless technology would be a stunning advance. While I don't understand the technology, and it may be pure BS, who knows, we'll see, but Rearden has a history of delivering on their claims, so cautious optimism is enabled.
An interesting bit for our purposes is how DIDO uses the cloud:
DIDO is a cloud wireless system. All of the intelligence of the DIDO system is in a DIDO Data Center, which then communicates to all of the users at once through all of the APs at once. So, you can think of the DIDO APs as a vast random array of antennas extending out from the DIDO Data Center for miles, but instead of running long wires from the Data Center to the antennas, DIDO uses the Internet to connect to each DIDO AP, allowing each DIDO AP to be placed anywhere there is an Internet connection, whether indoor or outdoor, much like a Wi-Fi AP could be placed anywhere there is an Internet connection.
DIDO communication begins with the DIDO APs exchanging brief test signals with the DIDO user devices. By analyzing what happened to these test signals as they propagate through the wireless links, the DIDO Data Center determines precisely what will happen when it transmits data signals from the APs to users, and how the simultaneously transmitted signals will sum together when received by each user device. Then, the DIDO Data Center uses this analysis, along with the data each user is requesting (e.g. video from a website), to create precise waveforms for all of the APs that, when transmitted at once will sum together at each user device to create a clean, independent waveform carrying the data requested by that user. So, if there are 10 APs and 10 users all within range of each other, then 10 radio signals will sum together at each antenna of each user’s device to produce an independent waveform for each device with only that device’s data.
From the perspective of linking the world's compute resources into one unified application fabric, this is perhaps the most interesting claim:
Cellular systems have long range, but have very high latency (up to 150 milliseconds or more for 3G). WiFi systems are fairly low latency (a few milliseconds), but are limited in range. There is no inherent reason for long-range wireless networks to be higher latency than wired or fiber networks, but the complex wireless protocols and distribution systems used in multiuser wireless systems add a great deal of latency. Even LTE (4G) incurs 40ms or more of latency Cellular systems have long range, but have very high latency (up to 150 milliseconds or more for 3G). WiFi systems are fairly low latency (a few milliseconds), but are limited in range. There is no inherent reason for long-range wireless networks to be higher latency than wired or fiber networks, but the complex wireless protocols and distribution systems used in multiuser wireless systems add a great deal of latency. Even LTE (4G) incurs 40ms or more of latency.
DIDO is extremely low latency, whether used short range or long range, because it can provide an independent channel for each user without having to rely on complex sharing protocols. In urban and suburban applications, DIDO will typically incur less than 1 millisecond of latency. DIDO Rural (using NVIS) incurs about 2-3 milliseconds of latency.
Here are some rough notes from the video portion on DIDO:
In 1948 Shannon up with the data capacity, the number of bits per second, that can be carried on a channel. This describes the limits on our wireless connections. The problem is we are running out of useable spectrum. Wireless is a shared space, we are all on the same channel. Only a small fraction of users can use the channel.
Shannon is correct, but he asked the wrong question. Rearden has come up with a completely new way to do wireless. It's different than how Marconi did wireless. It's different than how cellphones work today. They had to create new radios. It took years to make. They don't modulate the way any radio has before. They don't use wireless energy. They went back to fundamental principles of wireless propagation to make it work.
Using this new technology shared spectrum is not limited by Shannon's law at all. They can get 10x what Shannon's law is possible. They have 10 radios working, all at the Shannon limit on the same frequency and they do not interfere with each other at all. They think they can get to 100x, 1000x, they don't know what the limit is. They can do 100x what cellular does now with the same spectrum. This means no more dead zones. No more drops.
They are simple radios. Much faster than current wireless and lower latency. Can transmit up to 250 miles. This technology doesn't need towers anymore and it doesn't require a backhaul network. Radios are very cheap so can be included in lots of devices. They are seeing submillisecond latencies.
It's a whole new world. Started from the ground up building radio. We think of radio waves as invisible wires in the air, but they are not, they don't propagate that way. It's different. It's not TDMA, CDMA, or anything else.
Truly revolutionary if it works. If only the US had a telecom industry that was interested in revolutions...
The no backhaul contention doesn't seem plausible though. It takes time to hop between nodes in a P2P network, which is how I imagine their system would work. A fiber backhaul would seemingly yield much lower latencies over longer distances.
I'll also add that if this technology is really inexpensive, the most exciting advance would be in machine networks. Currently we are completely hamstrung in machine-to-machine communications with expensive technologies. We need a long distance, low power technology with dirt cheap components. If that technology could be found, it would revolutionize distributed sensor and device networks.
Benjamin Button, Cloud Everything and Why Shannon's Law Isn't
This is the video of the talk Steve Perlman gave at the Columbia Engineering School.
A major plot point is his observations on the sorry state of the VC industry. He asserts there's not much money for basic technology research anymore. VCs are a class of investment now that expect a return in few years using existing technologies. Corporate R&D is no better. Universities do basic research, but aren't product focused. So he founded a research incubator, Rearden, to do basic research. With Rearden the idea is to do risky research and make a hit that could change everything.
Along those lines is his plea to Reinvent Invention. I found these parts of his presentation really compelling. We have a problem of incrementalism today. The technologies today are so powerful people can make incremental improvements and make money. That's not what's needed. We used to have VCs that funded radical technology, that's no longer true. That needs to change. First to file patents will crush invention in the US. Invention makes the impossible possible. It's such a joy to make the things you work on finally work. That's why he became an engineer, to make things. If you are not taking risks then what's the point? Now that limited partners are organizations like pensions, they are risk averse.
Next is a large section on the real-time human rendering technology behind Benjamin Button. He goes into a lot of depth on OnLive, a cloud based gaming technology for playing video games over the Internet using a new video compression technology. Then he goes into DIDO, the wireless network addressed at the beginning of this post.
Let's Open this Up
I can't help but think Perlman's vision might have a better chance of succeeding if it were more open. By harnessing the enthusiasm and support of the maker and hacker cultures out there, he might achieve even more amazing things.
Currently everything Rearden does seems pretty closed. Their amazing graphics capabilities are accessible as a custom service. It took me about 10 minutes to find the developer link on OnLive and even that said email for more information. Perlman said there are something like 61 games for OnLive. Why aren't there 1000s? How many apps are in the Apple App store? Companies like Apple, Twitter, and Facebook are building large valuations partly by harnessing the power of open, social, and virality. Look how Microsoft's Kinect took off once it became a hackable open platform.
The reason why this even matters is for DIDO to really succeed it needs to be ubiquitous. It needs to be dirt cheap, it needs to be in everything and everywhere. For that to happen it requires a degree of openness on a completely different scale that has been been in evidence.
And part of what makes me concerned, strangely enough, can be found in the name of the company. This may be a bit of silliness, but it fits with the closed theme above. Rearden is the name of a main character in Atlas Shrugged. The plot of Atlas Shrugged has a lot of people who are very pleased with themselves picking up their toys and moving to an isolated location until the rest of the world finally breaks down and realizes their wonderfulness and begs for them to come back. It's like a reverse timeout.
I bring this up because of DIDO really works it could make an amazing difference in how we build things. Networking is by far the weakest part of our stack. Telecom moves at glacial speeds. We need disruption. I would hate to see advancements in this area messed up. Apple can be a do anything they want company because they make consumer goods that are so compelling people want them despite various Applisms. Rearden works on more fundamental technologies, and that's more about building connections, standards, and relationships and driving adoption. The Apple strategy won't work.
Maybe there are other ways to incubate greater success?
Another theme I loved in the talk were his examples of how people react to radical new inventions. Interestingly they either react with disdain or enthusiasm, nothing else in-between. Perlman's point is people consistently say things can't be done. And they are wrong. Go do impossible things.
- Part 1 of 5 Steve Perlman The Process of Invention
- Patent Application
- On Wired
- On VentureBeat
- On OnLive
- On YouTube
- On Hacker News
- How Will Memristors Change Everything?