Hyperscale is getting increasing amounts of press as the latest in a series of megatrends to rock the foundations of IT. Really a corollary of Big Data, and like it a radical new approach mandated by the demands of the huge data centers on which e new Internet giants run their mega-services, hyperscale is an answer to the issues of hardware component management in very large data centers. The problem is that when data centers grow beyond a threshold size, the brute force strategy of normal IT organizations of having technicians running around dealing with the plethora of individual component problems that constantly crop up becomes untenable. At the same time, the issue of vendor & version sprawl -- the "we have one of everything" syndrome --becomes unmanageable.
To solve those problems and bring hardware capital costs down to an acceptable level, the Internet giants have been building their own bare-bones commodity servers and arrays. Instead of putting management into each box, they have moved it up to an overall management layer, a software-led approach that let's them manage the entire data center as a unit. When a component such as a disk drive fails, they simply shut it down. When enough disk drives die, they replace the entire array. The hardware is not valuable enough to be worth repairing.
This architecture, universal among the Internet giants, is already creeping into the largest corporate data centers. It will become more common particularly as infrastructure for internal Big Data systems that will demand very large amounts of processing and storage. Once in those data centers its low cost will inevitably drive it into mainstream operations in increasingly software led, virtualized operations. Basically once data centers virtualize their entire environments, the expensive, feature-rich hardware from the traditional vendors is largely wasted, and ITOs can cut their capital costs dramatically by turning to bare bones commodity hardware and moving the control layer higher in the software stack to create the "single pane of glass" management of the entire data center.
HP's Moonshot announcement of early April is a recognition of this trend. While HP execs talked about selling to the Internet giants, it really in large part is an attempt to defend its position in the enterprise server market from the coming assault of commodity hardware vendors. It means that the HP server division will have to transition to a commodity financial model over time and learn to survive on razor-thin margins in a market where price, rather than features, is the differentiator. And HP as a corporation will have to transition to more of a software-led company, looking to its software and services divisions for its main sources of profits. But better that than losing a major percentage of its existing hardware market entirely.
What isn't being discussed however is the organizational impact of hyperscale. Because hyperscale is a direct attack on the traditional ITO, only one step down from the threat of complete IT outsourcing to the cloud. In a hyperscale data center, the small army of tech experts in servers and storage have nothing to do. Networking still has a place because those IT services have to be delivered to end-users, but software led networking from companies like Nicera (now part of VMware) the networking group will face its own obsolescence. And the virtualization of traditional servers and arrays will have a similar effect, eliminating most of what the technicians in those verticals do by moving all the controls out of the individual boxes to a higher, overall software layer and out to the edge of the architecture while automating most of the low-level optimization, scripting and similar activities. The result will be that a single person per shift will handle the work that a group of technicians were doing pre-virtualization.
Cloud computing also presents a large, medium-term threat to the traditional ITO. The writing went up on the wall when Salesforce.com grabbed a large chunk of the CRM market by going direct to the VPs of Sales with the message that they could have the 360° view of customers they needed now as a utility delivered out of a network plug in the wall just like electricity, without waiting two years for IT to build a system that never delivers everything the sales force needs at huge expense. Yes today IT is still for the most part safe, but sooner or later those legacy systems will fall, one by one, and the replacements will be delivered over the network. And even those huge Oracle and SAP systems are beginning to migrate to colocation centers and IaaS services where someone else runs the infrastructure.
So what will the ITO of the future look like? First, it with be much smaller. Second, it will be business oriented. Its main functions will be to design, locate and manage IT services that one way or the other will be delivered as utilities. Some may be owned by the enterprise but run in a colocation or IaaS center, others will be SaaS. Many will be services that do not exist today such as Big Data-based systems, that could about be run in today's enterprise data centers. They will be absolutely vital to the business of the next decade. IT's most important and most exciting function will be to identify how the enterprise can leverage those new services to gain competitive advantage. While that will sometimes be obvious, the more creative and in obvious the solution, the more valuable it often will be as a source of differentiation. What it won't be is about writing scripts or optimizing network switches.
Action Item: CIOs need to start planning for this future today. This means developing that menu of business services that IT delivers and reorganizing the IT budget around them. It also means assessing those services: Which should be retained in house and which farmed out to SaaS? Which should be moved out to a colocation service or IaaS vendor and which kept in the data center? What new services should IT contract from the cloud to do things like Big Data analysis that never were part of the IT infrastructure? It also means assessing the IT staff and starting the reorganization towards a services-led organization. The good news is that CIOs who start now have time to manage this transition. The bad news is this cannot be put off any longer. CIOs who ignore this revolution will be run over, and their ITOs will be road kill. Like it or not, we live in interesting times.