Proprietary data centers were the castles where firms guarded their treasures. These facilities were built to withstand any breach so the bank's technology infrastructure and precious records would be controlled and safe. While expensive, data centers allowed management to sleep at night knowing that the firm's processing infrastructure was safe, and its processes, trading technologies and critical data were beyond compromise.
Because these environments were hermetically sealed, virtually all critical technologies were managed and stored in the glass house (i.e., the data center). When rogue teams developed and ran mission-critical technology in uncontrolled environments, corporate IT and internal audit teams would swoop down with extraordinary zeal to capture and secure those environments in the glass house where they could be managed and controlled.
Then came electronic trading.
As speed became more critical, firms began to colocate their trading technology in exchange-owned or exchange-proximate data centers. While that isn't a heretical idea, shifting trading applications from corporate data centers to third-party data centers actually is a very destabilizing event.
First, this is not just any technology moving outside of the corporate-controlled technology environment - this is some of the firm's most sensitive infrastructure. Second, along with a firm's trading infrastructure, other market data, analytics and processing technology will follow. Third, and most significant, as more infrastructure moves out of the corporate glass house, the economics of a proprietary data center change.
Corporate data centers operate on allocations. Glass house real estate, electricity, cooling, processing, storage and support are allocated by the number of servers, processing power, storage and bandwidth used. The greater the resources employed, the greater the allocation. And this allocation can be very expensive.
The only way to reduce the allocation is by moving the technology outside of the data center - and that is exactly what happens when firms use colocation services. However, just moving IT from one data center to another doesn't reduce the cost of the fully owned data center - it just gets reallocated to the groups that stay.
If a significant amount of technology gets shifted, the economics become overbearing. Think of a firm where 10 percent of its technology moves from internal to external facilities. While the group that housed that 10 percent needs to pay for external space, the cost of that internal 10 percent that was abandoned in the corporate data center still needs to be borne by the firm and typically is reallocated to the other 90 percent of the users that didn't leave. That means the remaining businesses' data center allocations will increase by 11 percent (10 percent/90 percent). Ouch.
This will incentivize other groups to leave until either corporate halts this exodus or the last guy standing throws in the towel, as he is stuck supporting 100 percent of the data center allocation. It's like being the last taxpayer in a failing city - the last person who can afford to pay taxes supports the whole town. Talk about an impetus to move.
The process of migrating the corporate data center away from metro New York for resilience and cost, paired with the need to reduce latencies, is driving firms to colocate their most sensitive trading technologies outside of the corporate data center. This has put in motion a process that - combined with increasing robustness of software as a service, greater vendor hosting options, the increasing desirability of cloud computing and firms' desire to migrate fixed costs to variable costs - will spell the demise of the corporate data center.
While this process will not happen overnight and likely will take place over the next decade, the die is cast. And unless speed somehow becomes less important, or the economics of running a corporate data center change, owning this castle may not be the most economically sound business decision for the future.