Datacenters: The Future Is Not What It Used To Be
Did you see the New York Times article "Google Joins a Heavyweight Competition in Cloud Computing"?
Did you go beyond the headline and read how (paraphrased here) the industry is experiencing a "big shift" from the days when knowing how to build a datacenter was just as important as creating the right products?
The datacenters that companies had or had inherited by 2012 were born of a different computing era, when revenues were always growing.
This statement is highly controversial. Historically, datacenter management has been a core competency for large financial institutions. Those days are coming to an end. Moving forward, the effective management of computing resources will trump datacenter management as a core competency. As the old saying goes, "The problem with the future is that it is not what it used to be."
To start, let's consider recent events and the core drivers behind this trend:
- Fixed costs: The technology economics of datacenters are a massive fixed cost at a time when economic agility is needed.
- Slow growth: With the changes in financial services, we are no longer in an era of continuous growth in which large banks can plan for ever increasing computing resource needs, the big driver of datacenter expansion.
- Excess capacity: The downturn in many financial services businesses (e.g., trading, retail, home loans) has decreased the need for some aspects of computing capacity. Many firms have sold off some of their datacenters as they have become IT "empty nesters."
- No shareholder value: The cost of datacenters themselves ($200 million to $400 million a year or more) is a drag on the creation of shareholder value. For the large banks with 2 billion to 4 billion shares outstanding, this "drag" is worth 10 cents a share.
Couple the economic drivers with new demands from the business, and you have a formula that makes datacenters look somewhat — if not very — unattractive. For instance, consider the following emerging needs of the new business environment:
- An economic model based on "pay-as-you-go" services to provide economic agility.
- A service model based on "on-demand" fulfillment of technology services in real time.
- A "user self-service" delivery model and the disintermediation of IT middlemen.
- A transport model based on "portability and standards" in which businesses can have workloads processed at maximum value.
Now also consider the "new math" of computing power given the advent of cloud computing. But for this exercise, don't consider the public cloud vendors such as Amazon Web Services or Microsoft Azure. Let's just consider the enormous computer power within the world's top 10 banks. Here we have a platform of about 2 million mainframe MIPS, close to 1 million physical servers, and likely 1,000 PB of distributed storage capacity. At any point in time, approximately 40% of this capacity is not being used and likely 15% of it is now "excess" as a result of contraction.
The magnitude of this compute capacity could provide a massively scaled platform just for the financial services sector, with a specialized private cloud replete with the required levels of security and risk management to meet needs of regulators globally. Perhaps this is the future datacenter. It would be the banking datacenter of datacenters.
In the aforementioned New York Times article, David Campbell, the CTO of Microsoft's cloud group, said he was often surprised by the growth of the cloud. The article goes on to say, "And this shift is just beginning, maybe three years into a 10-year process that is democratizing heavy-duty processing power."
Maybe the future of the financial services industry datacenters is in those datacenters themselves. The core competency wouldn't be running and maintaining them, it would be actually using them for the business.
Eliminate Drag Dedicated datacenters in 2013 are a drag on a company's technology economics. The largest banks with revenues of $30 billion and up today have up to 2 million square feet of datacenter space. They may consume as much as 80 megawatts of power and run the facility at a cost of $200 million to $400 million annually. In turn, as scale diminishes, so does datacenter scale economics. All this does not portend well for the future economics of the datacenter.
In addition, the way technology can be delivered has changed drastically, as have the needs of the business. Today, technology directions take the form of IaaS, PaaS, and SaaS. Couple these with "cloud" and it's evident that future options for "compute" will likely be a mix of private cloud, managed private cloud, hosted cloud, and community/public cloud. It's conceivable that given this mix, only 20% of a firm's resources need to be in its own datacenter. Some models suggest that 100% of a corporate datacenter could "go away."
Does all this mean the end of the datacenter? More likely it points the way to a significant transformation of datacenters. The datacenters that companies have were born of a different era.
This next era of computing is characterized by the need for agility, innovation, globalization, and all-out digitization — the manifestation of technology, not just in IT but in all forms of products and services.
As such, the datacenter will take a new form. It may not have a physical address. The bank may not own it. But it will still be out there. Dr. Howard A. Rubin is a Professor Emeritus of Computer Science at Hunter College of the City University of New York, a MIT CISR Research Affiliate, a Gartner Senior Advisor, and a former Nolan Norton Research Fellow. He is the founder and CEO of Rubin Worldwide. Dr. Rubin is ... View Full Bio