I just came across a news article titled “GE preps industrial-strength cloud of its own“. At first glance, I figured here we go again, yet another company building its own cloud and boasting that it is the best solution for you the consumer. But then as the page loaded and I read the tag-line/summary, I actually got a little more intrigued.
“Today’s clouds aren’t optimized for heavy-duty industrial applications, says GE. Which is why it’s building a cloud of its own.”
Now this got me thinking… this isn’t your usual press release of yet another company going about how we are building our own cloud for our products; well actually it is, but it has a twist to it. GE are actually claiming that they will be building this cloud for the purpose of serving (specifically) industrial applications/systems.
So on Tuesday, GE is formally announcing Predix Cloud which it is building specifically to handle the types of data generated by jet engines, MRI scanners, power generation equipment, and other heavy-duty gear…
The interesting thing here is that this isn’t just any other cloud being built. For starters it’s being built by one of the behemoths of the industrial sector, and second of all this is a very purpose-built cloud that they will be building, i.e. not for everyone. Comparing this to Amazon’s cloud offering (Amazon Web Services), where you can just create an account, and start spinning up VM’s for general-purpose processing, hosting web sites, mail servers, online storage, or pretty much anything your heart desires. The way I see it is that this will not be the case with GE.
This bit of news then brings me to another thought… are we going to start seeing the rise of a new form of cloud infrastructure? The race so far has been between OpenStack, AWS, Azure, etc. But all these big players are competing in a specific arena, which is the public cloud, and more so, the general-purpose cloud. Since data of different industries and different applications behaves, and is treated in different manners, are we going to start seeing more companies build their own clouds, for application-specific purposes?
I believe this thought process is mainly attributed to the parallel that I see in the historical trend of microprocessors. You have your general-purpose CPU, such as the one in your computer right now, executing all kinds of tasks, as well as running your operating system. Then you have your GPU, which is a purpose-built processor used primarily for processing graphics instructions, visual transformations, and vector mathematics. Even more specific than that are Application-Specific Integrated Circuits (ASICs), which are pure integrated circuits that are built to solely perform a specific application (and damn are they fast). One very common occurrence where you may find ASICs is in a high-performance network switch or router that uses “hardware acceleration” technology to provide you with the specified performance; this hardware acceleration component is usually an ASIC chip that runs a certain set of procedures of an algorithm in near-real-time.
The trade-off between an ASIC and a CPU is that the former performs a limited and not very mutable set of instructions REALLY FAST, while the latter will perform almost any instruction (or set of instructions) that you throw at it, but it will take it’s sweet time (relatively) to do so.
So are we seeing the same trend begin to emerge today in the cloud infrastructure landscape? Is this what is to be expected to come out of GE’s new cloud, being that it’s a purpose-built cloud, and not just another general-purpose cloud like most of the others out there?
Your thoughts are welcome.