The Internet of Things is coming, and drastic traffic growth is going to blow your network sky-high. Should you scale up your on-premises data center? No. Should you move to the cloud? No. The best strategy, according to a speaker at the Interop 2015 conference, is to move your servers, applications, and data into your own servers in a top-tier colocation facility.
That’s not the advice you’d expect to hear in 2015, when the industry message is relentlessly cloud, cloud, cloud, cloud, cloud – and when the Interop expert speaker, Jason Mendenhall, carries the business card title of Executive Vice President, Cloud. However, when you realize that Mr. Mendenhall works for Switch’s massive 1.6 million square foot colocation center in Las Vegas, his bias toward colos becomes clear.
The bias becomes even more evident when the colocation data centers Mendenhall recommends have massive power and cooling, 100% uptime over the past five years, and massive redundant connectivity driven by 50+ telecommunications providers. Such as, by an amazing coincidence, the huge Switch data center in Las Vegas.
Let’s be clear: Listening to an exec from a colocation center talk about the disadvantages of the cloud for handling IoT traffic growth may seem like listening to the owner of a taxi fleet dissing Uber and Lyft — a message that’s on the wrong side of history. Not so: Large-scale colos are thriving. That’s true not only for private data centers like those operated by Apple, Google or Microsoft, but also for open colos operated by companies like Switch, Rackspace, Verizon, and others.
After all, the public cloud (supported by services from Amazon, Microsoft Azure, or Google) is not always the best answer to every IT question, especially for the very largest enterprises.
Mendenhall’s Interop talk, entitled, “IoT’s Potential Impact on Your Data Center,” kicked off by observing that we’re only at the start of IoT traffic growth, especially when it hits the enterprise. This is an emerging technology, not an emerged technology.
Today, there are truly few IoT applications that hit most corporations, though Mendenhall pointed out that United Parcel Service (UPS) has sensors that record 200 data points continuously about the positions, movements, routes, inventory, and delivery schedules for 80,000 trucks. Based on real-time data about traffic, weather and other factors, UPS can use Big Data from those sensors to change routes in real time, which Mendenhall says eliminated 206 million minutes of idling last year, as well as saving 1.5 million gallons of fuel.
Therefore, Mendenhall said, if IoT isn’t impacting your enterprise now, perhaps it will soon. The impact on a quiet data center could be significant in terms of scalability to support growth, density to handle the technology within a defined space, efficiency to sustain and conserve resources, bandwidth to enable connectivity, and proximity to other services (like Oracle or Salesforce) to promote collaboration.
Bashing the Cloud
What about the cloud, cloud, cloud, cloud, cloud? According to Mendenhall, the cloud is a fine place to start provisioning IoT services, but as the amount of data scales dramatically, it can become cost-prohibitive to rent the storage. “You want to make sure you own your data,” he said, claiming that about 50 terabytes is the crossover point. “Especially if there’s not a lot changing, you should own the infrastructure and run it yourself,” he said. “The monthly cost of moving the data into and out of the cloud can be expensive, especially if you want to keep data close to applications for real-time analytics.”
More arguments against the cloud to handle your IoT applications, according to Mendenhall:
- Limited hybrid cloud capability
- Some cloud providers might enter your business and become competitors
- Performance is lacking
- Cloud services start cheap, especially as you are starting up, but gets expensive
- It’s a black box; you don’t really know how it works
- Costs of connectivity can be unpredictable
- While you can export data, workloads are not portable between providers
- And once you scale up, you are essentially locked in.
Another option, of course, is scaling up your own on-premise data center. Yes? No. Mendenhall says that it’s a bad idea, because your core expertise is probably in your specific business, not in running and scaling a big data center. “You’ll have to upgrade your power, your cooling, your infrastructure, and your connectivity” in order to handle the workload, he says. He cited a large (unnamed) pharmaceutical customer that moved to Switch’s data center because they had so many telecommunications options. By contract, Mendenhall said, most data centers would be lucky to have two or three communications providers, and the switching costs could be prohibitive.
When I asked Mendenhall if he had any thoughts about how the IoT would potentially impact a data center beyond increased traffic, and if he had any advice beyond that enterprises should move their applications and data to a big colocation data center like his, he couldn’t think of any.
There you have it: According to a major colocation provider, the best solution for the hyperscale growth of the Internet of Things is to move your data and applications over to a major colocation provider. Forget the cloud, he says, because it’s too expensive; forget growing your own data center, it’s too difficult. That’s his story, and it’s a good one.