Forecasting is tricky, isn’t it?
Just ask Morgan Stanley's chief pontificator, Mike Wilson. At the end of last year Mike forecasted that the S&P 500 would drop to 3,000 points in Q1 of 2023. Yet here we are, heading into May, with the benchmark index steadily cruising above 4,000. (Mike shoots! Mike misses!).
Fortunately, the communications industry is defined by several trends that make predicting where it’s headed significantly easier than calling the stock market. And in the case of data centers, all of these indicators suggest that we will eventually see an inversion of the way the cloud is housed today.
Let’s begin with the basics: Data centers contain the computers that host the software that makes the cloud run. Today, it makes sense to centralize those resources in one place because the computers need to be able to collaborate on doing their cloudy business, and the shorter the distance between computers the faster the data can move between them (see: Steve’s First Law of Networking).
Keeping all of those computers in one building also makes it easier and more cost effective to manage them, power them, keep them secure and cool them down.
Want to learn more about cloud data center strategies? Register for our Cloud Data Center Strategies virtual summit today.
But there’s a paradox: co-locating the computers means moving them further away from users and businesses, which introduces network latency, which slows down the cloud apps that they want to use. In an ideal world — or cloud — the computing and the apps would sit next to the users.
Telcos and hyperscalers have recognized this, to a degree, and are installing so-called micro-data centers out at the cloud network edge. But some of these products fill a 40-foot shipping container, which stretches the definition of the term “micro,” and even the smaller ones still depend on a hoofing great data center sitting somewhere in the cloud network to offload some of the management and compute tasks.
Enter the data center diaspora
Change is coming. Within the next decade, three technologies will mature sufficiently to be combined to enable the creation of distributed cloud architectures where compute resources are entirely dispersed throughout the network – thus doing away with the need for giant data center redoubts containing serried ranks of server hardware.
In this new world, cloud networks won’t build data centers, they’ll build a data diaspora.
The three technologies that will make this transformation possible are:
1. Next-gen computing
Computing technologies that dispense with transistors will provide the foundation of the data diaspora. Once they reach production readiness, they’ll make it possible to shrink the compute power available today in a 42U high, 19-inch-wide rack-mounted server system down to the size of an Alexa speaker. Quantum computing has been getting the lion’s share of attention to date — partly because of a dire Marvel movie — but keep an eye out for developments in Graphene computers and Nanomagnet computing, also.
2. Ultra-high-speed networks
In the ‘80s, the standard unit of bandwidth for building corporate networks was a T1 leased line running at 1.55 Mbps. Comcast Xfinity now sells consumer service running at 6 Gbps, almost 4,000 times faster. Yet, the comms industry is still nowhere near maxing out the throughput possible over fiber-optic cable, with both terabit and even petabit speeds being demonstrated in labs (Developments in Probabilistic Constellation Shaping are enabling these mind-blowing speeds).
General availability of ultra-high-speed connectivity will ameliorate the need to install servers in the same building, let alone the same rack. (Steve’s First Law of Networking? More of a guideline now).
3. Time travel
Oh, wait, sorry, nope - wrong article. Hang on... Ah, here it is...
3. Autonomous automation
As networks have grown in complexity and importance, demand for automation solutions that can increase efficiency and head off network issues has increased. According to Gartner, 35% of enterprises currently use some sort of automation capability, but the shift to a data diaspora will grow that to 100% as the scale, speed and sophistication of the coming distributed cloud infrastructure negates the ability for humans (a.k.a. puny mortals) to participate actively in their management.
Instead, oversight of the diaspora will be handled autonomously by automation code in conjunction with advanced AI. In terms of network management, people will just be along for the ride. Their job will refocus around making top-level strategic decisions relating to business outcomes (SLAs, pricing and so on).
For both telcos and enterprises, this will require a change in culture as much as an upgrade in tech. Network architects are notoriously reluctant to give up any control of network oversight. Makers of advanced automation solutions will need to convince them that they can put their products in full self-driving mode before they go to bed, and not wake up in the morning working for the robot insect overlords.
Winners and not winners
The inversion of the data center model may be difficult to envision today, given the current rampant growth in both the number and size of data center facilities around the world (China Telecom holds the record for the largest data center in the world in Mongolia, covering an area the size of 185 football fields).
But keep in mind that communications is an industry defined by contrarian outcomes; where sh*t doesn’t get real, it goes virtual; where it turns out that bigger isn’t actually better - smaller is the superior size.
Compared to previous communication transitions — the switch from analogue to digital, say — a restructuring of “where the boxes go” in cloud infrastructure isn’t really that revolutionary.
Which is not to say that it won’t have a profound effect on the fortunes of some companies involved in provisioning cloud technology and services.
Vendors that combine a proficiency in making very small, very intelligent comms devices, and also have a deep understanding of both cloud and telco edge networks, should clean up from the advent of the data diaspora.
At the top of that list are Huawei and Samsung, maybe AWS — but not Apple.
Market leaders in advanced automation/AI will also benefit, including Ciena/Blue Planet, IBM, Itential and RedHat.
On the other side of the ledger the prognosis for traditional data center colocation companies looks decidedly gloomier. A big part of the problem for companies like Coresite, Equinix, Digital Realty and Rackspace Technology is that they make so much money from renting out conveniently located data center real estate that there hasn’t been a pressing inducement for them to get innovative with higher-level network and service stuff, which they tend to view as their tenants’ thing, anyway.
But in a world where data centers are being disintermediated and/or distributed, providing 24x7 physical security, redundant diesel generator power backup, and a really big AC unit will no longer deliver the value they once did.
Still, the good news is these companies probably have a while before they have to convert their facilities to mushroom farms, paintball arenas, homeless shelters or even go kart tracks.
How long? I’m not going to predict exactly when all of this is going to happen because, for one thing, I don’t know (exactly). And for another, unlike Morgan Stanley’s Mike Wilson, these days I am trying not to be that know-it-all blowhard who enjoys taking frequent gulps from his frosty tankard o’ made-in-America hubris.
That said, there is no question that cloud network architects that want to create a successful cloud strategy for their business should be keeping a beady eye on the emergence of the data diaspora.
Want to learn more about data strategies for the cloud? Register for our Cloud Data Center Strategies virtual event here.