08 August 2016
I was a teenager in the fall of 1984 on a typical hot, dusty afternoon in Pilani, India when I walked into a data center for the first time -- and it was love at first sight. In front of me was an early computer, an IBM 1130. First sold in 1965, it was already an ancient relic by then. Nonetheless, I was mesmerized by the blinking lights, the rows of switches and the clattering punch-card reader, so much so that I was oblivious to the shock of the 30-degree temperature drop I had just experienced.

Data centers have come a long way since the days of the IBM 1130. However, their purpose is still the same. The purpose of data centers hasn't changed since Alan Turing helped build the world's first one in Bletchley Park, UK in 1943. He had to start by building the world's first programmable, electronic, digital computer, the Colossus, to crack the "unbreakable" enigma machines used by the Nazis during World War II. They didn't call it that in 1943, but around the Colossus was, of course, the world's first data center. In fact, even before the Colossus, Alan Turing had already written his famous paper laying the underpinnings of modern computer science. From this paper we derive that all modern programmable computers are essentially equivalent and that it is actually possible to build a complete computer with just one instruction. So then, why do we build data centers when they are all, in effect, just running one instruction over and over again?

Data centers make magic possible

Data centers are built to process data, and it is their technology that makes the Facebook news feed so addictive, Google search so powerful, and Amazon shopping so lucrative. It is what allows Siri to "understand" what you say.

Originally, data centers, such as those running the IBM 1130, were built around one machine. Each machine was expensive and it took a lot of care and feeding by sophisticated people to keep it running. This evolved into groups of machines in the 1980s, but it was still essentially more of the same. It was a high-maintenance endeavor. 

Then came what was later to become Google, as we know it today. It all started with research in the 1990s at Stanford to scale a search engine - by not one or two, but three and even four orders of magnitude. They needed to create a new way to index data at a scale that had never before been considered, and it led to a reimagining of the data center. At the same time, VMware commercialized and popularized an old idea -- virtualization -- which loosened the bonds between software and hardware, making servers much more fluid. Since then we have seen data centers change dramatically. Those changes have occurred on several levels -- physical, conceptual, and in the marketplace.

The most obvious change in data centers can be seen at the physical level. The first data centers were glass, climate-controlled rooms filled with IT experts managing a single machine. These machines required a lot of administrators to manage and maintain them. The ratio of humans to machines flipped when we moved to minicomputers and then workstations that were easier to use and lower-maintenance. We assembled machines in racks in a cage and needed fewer administrators. A typical data center went from one to tens, and sometimes hundreds, of machines. Today, we have arrived at a physical layout where there are no humans to be found. Instead, data centers are comprised of vast rooms (and often warehouses) of machines that are completely maintained robotically. Modern data centers usually contain tens of thousands of machines.
The ratio of humans to machines flipped when we moved to minicomputers and then workstations.
Share this

Is your server a pet or is it cattle?

This physical change is symbolic of the conceptual change we have seen. All the work done in the original data centers was mostly self-contained. All the data, programs, and even people were physically working together in the same location. Data transmission typically consisted of a tape of data sent via mail or courier between data centers. Now with the advent of cloud computing, the work being done lies within a virtual idea of a program being run and distributed across the globe. Seemingly small, the magnitude of this change becomes even clearer when you look at how this has changed jobs associated with the data center. There's a well-worn phrase used in the IT world that cheekily illustrates this paradigm shift: "Is your server a pet or is it cattle?" The data center of old required high-touch management of and care for each box like a beloved pet, whereas today's data center consists of interchangeable computers, none of which are special. Unlike a pet, losing one causes no heartbreak. They can also be set up to maintain themselves, just as cattle are put out to pasture to graze on their own. The latter requires a lot less administration and, therefore, fewer people.

When the boxes changed, the leading companies in the space changed along with them. In the 1980s and 1990s there was a standard formula of equipment in every data center. Servers were HP, Dell or IBM. Networking came from Cisco or Juniper. Your database was Oracle, and your operating system was Microsoft. Most of these incumbents are nowhere to be found in the new landscape. It looks like Microsoft is the only legacy player who will continue to thrive, thanks to their now rapid acceptance and embrace of this new reality. They sit alongside Amazon Web Services (AWS), Google, Facebook and Alibaba as the new masters of the data center universe. These companies came in and built their own proprietary systems, often using open source, and are even beginning to build their own chips. They've leveraged these systems into massive businesses, where they rent out use of their technology to other companies that don't have the scale to build and manage it themselves. This is the so called "Cloud" (of computers), and it made computing tremendously efficient for everyone from startups and enterprises to governments. 

What drove these companies to make massive capital expenditures to reimagine and build new data centers? It all comes back to the fundamental quest to solve problems. Engineers build new technology by taking a question, turning it over, and looking at it from a different angle. The modern, super-powered data center removes the obstacles of scale and speed, and it ultimately lets engineers create previously unimaginable solutions to problems. It frees their vision and enables them to tackle any challenge. The future of self-driving cars, virtual reality, and IoT (Internet of Things) will all be created leveraging the vast scale and capabilities of these data centers.

What lies ahead for data centers? We can look at this both qualitatively and quantitatively. The qualitative future of data centers lies in their conceptual evolution. While there will always need to be a room full of computers, one of the most important changes is that a data center is now available on demand, as a service, for rent and use to anyone with a credit card and the technical knowledge to leverage it. The magic and power of the new data center is that it is an enabler for human imagination, and there are no limits to the solutions the human imagination can create. Just as the modern data center freed up tech workers to spend less time administrating and more time solving problems, the solutions they develop will bring us more freedom in our everyday lives. This same ability to automate will drive cars for us, make purchases for us, call home for us, and even turn off the stove for us when we realize we forgot to do it ourselves. These are just a few of the advances smart engineers are already working on. The real future lies latent in their imagination 

Quantitatively, the commercial reality of the data center market is hundreds of billions of dollar every year. This massive prize has attracted a plethora of entrepreneurs and startups. Innovation very often comes from startups, and the rapid and revolutionary changes we're seeing in data centers provides fertile ground for them. As the founder of Engineering Capital, I have the privilege of investing in and working with some of the leading entrepreneurs and startups in this space. This provides a unique vantage point, and based on what I'm seeing I expect several new, independent, billion dollar companies to be created from such startups.
The magic and power of the new data center is that it is an enabler for human imagination.
Share this

Massive prize has also attracted the behemoths

Big players have also been attracted to this market. Amazon AWS has leaped ahead and has an early lead in infrastructure. They have amassed enormous scale and rich functionality and, consequently, have established a significant advantage. AWS is focused on traditional IT and is looking to lift and shift existing workloads to the Cloud. 

Google, on the other hand, has powerful technology in-house, though they have not applied it to traditional IT. Their machine learning technology, in particular, is excellent, and they will attempt to leverage this as their competitive advantage. Google is focused on cloud-native applications, which are built from the ground up to operate not on one machine, but on 10s or 100s of machines in parallel in a distributed fashion.

I already mentioned Microsoft as the one legacy player who appears to be crossing the divide into this new market. They are leveraging their franchises in Office, SQL Server and SharePoint, any of which could be a silver bullet given their deep relationships with enterprises -- small, medium and large. 
Each of the incumbents sits on tens of billions of dollars in cash. They are not accustomed to losing, so I expect a fierce fight. Let the games begin! 
Tap to read full article