Picture this: there's a sensitive site at a remote location, and a bunch of data needs to be analyzed in a hurry. Maybe there's a potential data breach, maybe you just need some fast answers about a complex fault in a system. What's important is that you don't have the capability on site, but you need it there now.
There's no bandwidth into the site, so you can't just suck the data into a cloud of some kind and put your smart folks to work on it. Besides, the dataset is too big. It'd take days.
You could send someone in with a carton full of tapes—or maybe an AWS Snowball—but this isn't the kind of data you can just send off into the cloud; it's sensitive stuff, and you're security minded. Besides, that still takes ages to get loaded, and you need answers in a hurry.
What if, instead of taking the data to the cloud, you brought the compute to the edge?
X-IO has taken its data dense, high-bandwidth, low-latency Axellio hardware and put it in two travel cases that can fit on a plane.
Like something out of Mission Impossible, the appliance breaks down into two sets of parts: the chassis that goes in the hold, and up to 6 canisters of storage that stay with you as carry-on. Sensitive data need never leave your sight. When it arrives at its destination, it re-assembles in moments, ready to do its big data dance.
Your Tactical Applied Data Analysis team (I just made that up) of expensive data wizards can fly out to the location, do their magic, and then be on their way with either a raw dataset (the Axellio can hold up to a 460TB of data today) or a summary set of data to take back to headquarters for more leisurely analysis. In a world where fast answers to tricky problems needs some seriously cool tools, the portable Axellio makes easy what would have been tricky—if not impossible—before.
Gavin McLaughlin, Strategy and Communication at X-IO Technologies, told me that the product came about after a wishlist request from a defence customer, and engineering couldn't help themselves. "What would have taken a palette load of gear is now available in two suitcases," he said.
I can see a few immediate use-cases: security breach response, seismic analysis, on-site research, and X-IO has addressed some of those already. The Axellio supports offload modules such as Intel Phi for parallel compute, Nvidia K series GPUs for video processing, and P series GPUs for machine learning, plus the Solarflare Precision Timing Protocol Packet Capture offload.
But that's just the start. I can see a bunch of edge-computing use-cases where people would love to use cloud-style machine learning approaches but just can't get the data to it fast enough to be useful, if at all. All the sensor data starting to accumulate from the myriad of IoT devices that are proliferating as fast as the data they generate. Complex industrial control systems with tricky optimisation issues.
X-IO isn't selling this kit direct to customers, preferring to partner with folk like Ascolta, Solarflare, Levyx, and Issac to build systems on top of the Axellio hardware. It doesn't even have an operating system, so partners can load up whatever software stack they think works best.
Edge computing is really only just getting started, particularly now that cloud people have discovered that physical laws aren't merely guidelines and the speed-of-light is a thing that exists. I'm sure a bunch of people smarter than me will find all kinds of amazing things to do with plenty of compute capability out at the edge.
Now it just needs a catchy theme tune.
