Top-of-Mind for Data Center Leaders: Scale!
Scale is a top challenge for many iMasons, with demand reaching levels completely unimaginable just a few years ago.
IoT leader Stephen DiFranco has predicted a doubling of the amount of compute and network that’s going to be outside of the data center. Another data center executive predicted we’ll need 4,000 more hyperscale data centers over the next 3-5 years.
2x more compute and network. 4,000 more hyperscale data centers. That’s scale like we’ve never seen before. “The data center prediction could be off by a factor of ten and that’s still unimaginably huge demand,” said one iMasons partner.
No question about it, scale is a top-of-mind issue for hyperscale tech companies and the partners who supply them. (Not that it’s a worry for all; it keeps some up at night, but others see it as a technical problem that can be solved – given the right people.)
“No question about it, scale is a top-of-mind issue for hyperscale tech companies and the partners who supply them.” – Click to tweet
Scale was a key topic of discussion during our most recent meeting with the Infrastructure Masons Advisory Council (which includes five end users and eight partners). It was also the focus of a panel discussion that followed the Advisory Council meeting. The panel featured Dean Nelson, Christian Belady, and Joe Kava – all Advisory Council members. Key highlights include:
- How much scale? “Whatever size you plan to scale to, it’s too small.” “[The] amount of compute power [available today] is beyond what any of us could have imagined even just a few years ago. So who knows what’s going to happen in the next couple of years.”
“Whatever size you plan to scale to, it’s too small.” – Click to tweet
- Public cloud is a whole different ballgame. “When we were building data centers to house the compute infrastructure for our own products (Gmail, search, even YouTube) we were able to stay relatively ahead of the demand because we could look back at historic demand and build in a buffer. Those were the good old days. With the public cloud we’re building for everyone else’s products, which I have no idea about (and often they have no idea about).” (Proof point: Pokémon Go.)
- And that changes everything. “20 years ago it was completely implausible to conceive of a startup that went to 100 million users overnight. Today that happens. The business models that global connectivity has enabled have fundamentally changed the needs for building out infrastructure.”
“The business models enabled by global connectivity have fundamentally changed how we build out infrastructure.” – Click to tweet
What’s driving this scale?
Scale is driven by VR, AI, IoT, 8K, and other acronyms that essentially mean the same thing: more data, more compute, more distributed. It’s driven by the same kinds of trends that drove the rise of the cloud – relatively low-cost compute and network – and falling hardware costs, as well as the cloud itself, and the innovation driven by cloud players.
On the business side, “Business units and smart forward-thinking individuals say to themselves, ‘We have to deploy solutions.’ How? ‘Let’s run a cloud instance.’ And they put in place sensors to gather data. IoT is almost by default outside of the realm of the CIO and CTO. End users are embodying IoT even if they’re not thinking of it as such,” explained one partner.
“There will soon be 2x more compute done outside of what we call data centers.” – Click to tweet
That partner calls it the “bimodal distribution of compute.” He said, “There will soon be two times more compute done outside of what we call data centers – at the edge, distributed compute, devices. And that has to be done in places where there might not even be roads.” Most worrisome is the people aspect. “It is going to happen but who has the skill sets to invent, deliver, and keep it operational?”
Scale stresses the supply chain
For one end user, the top-of-mind worry associated with scale is its impact on the supply chain. “The scale is overrunning the supply chain. If you’re a supplier of anything infrastructure related, the demand is not going to abate. If you’re not caught up today you’re not going to recover while we [end users] continue to hammer on you. This goes for everything – even power is starting to look creaky.”
“If you’re a supplier of anything infrastructure related, the demand is not going to abate.” – Click to tweet
The end user added, “We want to do more specialization but we have to do less because we have to fix the supply chain.”
Solution? A collaborative industrial data center park
Advisory Council members talked about potential solutions to the scale challenge. One that rose to the top: collaboration.
Raising the idea, one end user said, “We’re all going to be massive in the future. Why are we doing things fragmented and driving infrastructure we don’t really need? What if we all go together to a country and, for example, take 2,000 acres and put a power plant and network and drive hyper efficiency on how we even use this stuff. That’s not even groundbreaking; it’s unprecedented.”
Might that put companies at a competitive disadvantage? One end user replied, “We build data centers out of necessity. It’s not what we compete on. At the end of the day we compete on services. We already partner on fiber so a collaborative industrial data center park isn’t that much of a stretch.” And yet, chimed in another end user, “It’s harder than it should be.”
“We build data centers out of necessity. We compete on services.” – Click to tweet
There could be clear economic benefits to the business, though. Said one end user, “In a collaborative effort we could do it so much better and drive so much efficiency. Then we’re a different player in town. Even at our current scale, being five times bigger would present even more opportunities than we have today.” Another end user agreed, adding, “Where there’s a single buyer it’s called a monopsony. The buyer for a 1000 MW power plant is us and an aluminum smelter. A monopsony could drive a lot of opportunities.”
“The data center is going to become a utility.” – Click to tweet
There would be hefty economic costs, though, too. Said another end user, “A 1000 MW power plant is about a $4.5 billion investment. So if you converge on a solution at 1000 MW, that’s a $100 billion investment.” He added, “You show me the investment, and I’m in.”
“There have been things similar to what you’ve described,” said one partner. “The problem is that the technology makes sense, but ultimately there’s a matter of concern about concentration – having too much critical infrastructure in one area.” True, responded one end user, “We have that problem in Virginia.” (Check out the resiliency study.)
But one partner sees the future clearly. “The data center is going to become a utility. Economies of scale work better in this business than in others. Building a 1000 MW data center is much more efficient than building a 100 MW data center.”
This is the last in our series of eight blog posts reflecting the top-of-mind issues discussed during the most recent Infrastructure Masons Advisory Council meeting. All 8 posts can be found at imasons.org/blogs.
Previous posts in the top-of-mind series
The Biggest Challenge of All? People
Our Environmental Legacy
Shifting Power & Shifting Priorities
Just How Critical Is Our Critical Infrastructure?
Emerging Markets Pose New Challenges
Alignment Between IT, Facilities & The Business
Never Stop Innovating
Infrastructure Masons was established to provide infrastructure executives and technical professionals an independent forum to connect, grow and give back. If you’re not already a member, learn the benefits of joining at https://imasons.org/join/.
Join the discussion