Computing steps closer to the network's edge
Connecting state and local government leaders
Edge computing offers the ability to conduct real-time analytics with less latency and greater scalability.
As governments take advantage of emerging technologies such as drones and come up with new ways to leverage established tech such as GPS, the importance of edge computing becomes more apparent.
Although it uses a notion of cloud computing, edge computing is different. Cloud computing consolidates many machines in a large data center, while edge computing enables processing closer to the source -- or at the network edge. Take wearable technology such as Google Glass or a hobbyist drone with a video camera. They are too small and light to carry enough computing power to process the data their collecting.
“The approach is to transmit the data in real time wirelessly, and here’s where the difference between using cloud computing and edge computing comes into the picture,” said Mahadev Satyanarayanan, who runs Carnegie Mellon University’s Living Edge Laboratory and pioneered edge computing. “If you try to send [the data] to the cloud, the nearest data center is going to be far away. It has to be. That is the whole point of cloud computing," he explained. But applications that call for "extremely timely response combined with very substantial compute on the sensors’ data is the combination for which cloud computing just doesn’t work. And so it is for that, edge computing is crucial.”
Instead of being transmitted to a distant data center, the data goes to a "cloudlet" situated nearby, such as in a vehicle also in the field, which reduces end-to-end latency. Cloudlets are small data centers -- a single rack of computers in a closet or a small box in a Humvee -- that use cloud computing features such as multitenancy and elasticity, said Satyanarayanan, who goes by Satya.
In the public sector, edge computing can make an impact in several areas. One is military operations. For instance, if a video camera on a drone captures what appears to be a target, the analytics to confirm that and the controls directing it to zoom in for more details must happen in real time, before the drone has flown past the potential target, Satya said.
The same applies to disaster recovery after a hurricane, earthquake or other event. Google Maps, whose data comes from the cloud, could show a driver that a given road is coming up, but a report from drone footage of the last vehicle that followed that path could show that the road is inaccessible.
“Data from the edge, from the field, is more important and in many ways dominates information from the cloud,” Satya said. “The data at the edge has to be processed in real time so you aren’t fooled by the stale data from the cloud.”
In addition to low latency, edge computing offers other benefits. One is bandwidth scalability. If a city has 1,000 video cameras on 1,000 vehicles and they are all livestream footage at the same time, the cellular infrastructure has to support wireless transmission of several gigabits per second. Doing some of the processing in a cloudlet uses less bandwidth.
Edge computing is also secure, Satya said. That’s because the data has less distance to travel, giving hackers less time to disrupt it. What’s more, he added, cloudlets could still be operational even when traditional cloud access is denied.
In Pittsburgh, where CMU is located, Satya runs the Living Edge Laboratory, which has positioned antennas around the city that are connected via fiber optics to a cloudlet in the lab. This means that signal from people's mobile devices in those areas can tap into the cloudlet. Satya is working with the city to set up additional smart city-type initiatives, but nothing has been deployed yet, he said.
Last month, the lab announced a two-year agreement with Microsoft under which the company will provide edge computing products such as the Azure Data Box Edge, an appliance that transfers data to Azure Storage from the edge; Azure Stack, which lets organizations use Azure cloud services in their own data centers; and Microsoft Azure credits, which provide access to cloud services such as artificial intelligence, the internet of things and storage.
The lab is open to CMU students and faculty in addition to companies looking to test their ideas, Satya said, and he looks forward to the results.
“The ability to process, to bring compute resources close to the edge to perform real-time analytics is very, very powerful,” he said.