In the final scene of The Terminator, the machine sent back from the future to murder poor Sarah Connor has had all of its oiled Arnie skin ripped off. It marauds through a facility full of factory robots, which point to the beginning of our end. And one of those robots was the same one IBM used to pick up and move tape back in the early 1990s.
That machine was originally used by General Motors to spray paint cars, but was adapted by the tech titan. Big Blue has continued to use robotics for its tape libraries, as has almost every other big vendor managing that old yet resilient storage technology.
But over the last two decades, innovation with robotics in the data centre has been almost non-existent. There has been little motivation from businesses and vendors alike to spend large sums of money on heavy automation. Meanwhile, the lack of standards surrounding rack design and technology, and an absence of modularity, which would make it easier for robots to actually tinker with hardware, has meant attempts at greater automation have been abortive.
But, thanks to the rapid growth of the cloud industry, more sophisticated robots are set to cover the world’s data centres. The key to winning the cloud race, according to Rackspace’s Nigel Beighton, is being able to scale at incredible speed, across millions of servers. That’s because of the “burst” model many customers are now demanding, where they where they can quickly push workloads out to the public cloud when usage spikes.
To really ramp up scale time, automation needs to rise accordingly. That’s why Beighton thinks robots will be found in many cloud vendor facilities in the coming years. “If the key to being competitive in the cloud space is being able to scale quickly, that’s why automation is really key moving forward. It’s the automation that is a key part of making you price competitive in this market,” he says.
This automation drive will partly be inspired by the Open Compute movement, led by Facebook engineers, as it creates the standards around modular architecture that robots need to operate effectively, he says. Indeed, for robots to thrive in the data centre, the innovation actually needs to take place in the rack, not in the robotics.
“The robot technology is there now … we see it all the time in warehouses and car factories, the ability for them to do precision stuff is there,” notes Beighton.
“It’s only the fundamental design of the components it’s dealing with that is the limiting factor. But that’s starting to radically change because cloud is demanding new types of server technology.”
Greater automation leads to greater efficiency, which should mean reduced costs and higher service levels for the end user. But what would robots actually do in the data centre?
We can look to IBM to answer that. It’s already built a mini robot army, which is busy patrolling data centres worldwide. IBM will soon have over 20 working on its own premises, with another five or so used by IBM consultants for customer facilities, Jon Lenchner, researcher at the IBM TJ Watson Research Center, tells Cloud Pro.
Most are used to carry out audits of the data centre. They map the facility, using the standard tiles and some odometry data to do navigation, with something “like a slightly drunken walk”, Lenchner says. They check temperature at different points of the data centre, from the floor to top of racks. All the data it takes is then fed into an IBM analytics engine, which supports cooling optimisation efforts.
IBM claims that on average this has helped realise 10 - 20 pc energy efficiency gains in its own data centres. “You almost always realise a return on investment within a year of that first scan,” Lenchner says.
But IBM is toying with other ideas. It is experimenting with RFID tags, attaching them to different assets across a complex. When something starts to fail, the robot could quickly locate the hardware for an on-site technician to address the issue. There have been some issues with picking up signals accurately, and with precision.
“It currently can’t tell you exactly what rack an asset is in, but it’s within one or two tiles, which is generally good enough. We think in one or two months we’ll be able to tell you exactly.” Lenchner says.
The next stage is to look at having the robots open a rack and play with the hardware. It has thus far proven tricky. “Opening a rack even for a human being is a little challenging - we haven’t tried to get it to do that yet,” the IBM boff adds.
When racks are more standardised, thanks to Open Compute or other open projects, the task will be simpler. IBM, which is using a Roomba base for its robots right now, could use something like Baxter, a friendly-faced robot that has claws for interacting with objects, as it looks to make its machines more functional.
“If we wanted to get into the business of replacing blade devices, we could employ the Baxter technology. Some day they might be the technology of choice, it’s definitely exciting to watch,” says Lenchner.
The day when bots can do both the monitoring and the rack actions will be the day they have finally made their mark on cloud and IT in general. We await our robot overlords.
