Unified data centre infrastructure

In the early and mid '80s, IT organisations typically bought their infrastructure from a single source.

“You went and got everything from one vendor,” says Dr Kevin McIsaac, analyst at IBRS. “You went to Digital and you bought a VAX, and that was networking, storage, servers, operating system, and database - and sometimes even applications - all pre-integrated from one vendor.

“The problem was, those siloed, vertically integrated solutions were generally quite expensive, somewhat inflexible and you basically had to buy into a whole series of choices, and that meant you couldn't do anything else,” he says.

Throughout the '90s, a new paradigm emerged, which McIsaac calls the 'layered component model': the IT department would look at each infrastructure vertical - storage, networks, servers, operating systems and so on - and select the best fit for their organisation in each layer. The IT department would then be responsible for tying all these pieces from multiple vendors together, and ironing out any kinks that cropped up through the life of the gear.

This model forced vendors to compete within each of these interoperable infrastructure stacks, resulting in several benefits for customers: it drove the down capital cost of the equipment and it forced vendors to constantly innovate within these infrastructure verticals in order to remain relevant.

So far, so good.

But now, in 2010, this system is beginning to weigh IT departments down. While the layered component model successfully drove down the capital costs of IT infrastructure, it has resulted in a data centre that is filled with disparate infrastructure from many vendors. Thus, integrating, managing and resolving conflicts inherent in this mismatched kit has become a significant burden on the IT department.

“People have started to realise that they're spending a lot of time acting as a systems integrator,” McIsaac says.

In fact, the layered model has worked so well that capital costs have become a small portion of TCO; more money is actually spent on running this infrastructure. Much, much more, according to Paul Robson, VP and GM of HP Networking, Asia Pacific and Japan.

“The analysts - Gartner and IDC - estimate that approximately 80% of a CIO's budget is spent on running infrastructure, and about 20% is spent on buying or innovating new infrastructure,” Robson says.

Why is this figure so big? Robson says: “80% of the spend is primarily spent on running multiple OSs, multiple vendor technologies and multiple architectures. Big businesses end up with multivendor environments, multiple operating systems, multiple management tools and a lot of complexity. And that complexity is expensive to run.”

The replacement

The new unified infrastructure model aims to address these complaints. The basic idea is quite simple: vendors will stitch together the basic building blocks of servers, storage and networking, and offer them as complete, all-in-one bundles of infrastructure. This means the IT department won't have to go through the rigmarole of integrating these layers when first deploying the kit.

Ongoing administration should be easier under this model, too; since these bundles typically have a single architecture across the whole thing, it's a lot easier to make changes to configurations, provision compute power and make upgrades without conflicts. In practical terms, this translates ultimately into lower operating costs - your IT team spends far less time managing your environment.

In his research note 'Return of the systems vendor', McIsaac notes that this trend is reminiscent of the integrated offerings of the '80s, but instead will “be founded on robust standards such as TCP/ IP, FC, NFS/ CIFS, x86/64 and Linux/Windows”. Therefore, we'll sidestep the problems of the proprietary '80s model, namely high capital cost and vendor lock-in.

The vendors involved have a few different approaches to this development. Some are going it alone; HP, for one, relies on its recent acquisitions - like 3com, Colubris, IBRIX, LeftHand and EDS - to provide the necessary tech for each infrastructure vertical. The company's Converged Infrastructure range of products combine servers, storage, software and networking.

Like many products in this space, HP aims to simplify infrastructure management. Robson says: “We have one window of management across those devices, with the same management tools, so that when you train an engineer or an end user to run your environment, it's the same management tool that runs the servers, the storage or the network. So you're training the guy once, as opposed to on three different platforms.”

The company's BladeMatrix product provides a simple example of the unified infrastructure model: it's a single device that contains server, storage and networking capabilities. The company also offers other devices that comprise various combinations of x86 servers, storage devices, networking and software.

Other vendors have banded together to offer unified infrastructure systems; the VCE alliance, for example, comprises VMware, Cisco and EMC. The trio's work has resulted in Vblock, a platform that integrates networking, storage, servers and virtualisation in a single chassis. Technology services provider CSC is offering infrastructure services in Australia from a pair of Vblocks.

Bob Hayward, CSC's Chief Technology and Innovation Officer in Australia, explains the technology: “[Vblock] dramatically reduces the number of different components in a chassis, avoiding duplication. It's really fusing together network controllers, network switches, storage controllers and blade servers in an integrated way with virtualisation built in at the design point rather than as an afterthought.”

Like others in this field, the Vblock also features simplified management: “[It has] single pane of glass management, in an integrated way, which makes the whole management of the virtualised environment - the cloud - simpler,” Hayward says.

*Andrew Collins is a freelance writer.