Quocirca's Straight Talking: Grid's a 'no-brainer'

In just three years' time, grid computing has gone from obscurity to mainstream acceptance, says Quocirca's Clive Longbottom. But that doesn't mean there aren't still a few issues holding back IT departments.

Quocirca has been tracking grid computing for some years now and our latest research report, carried out in conjunction with Oracle, brings up some interesting points.

Firstly, although there are geographic differences in how grid is viewed and how it is being implemented, the capability to move to a grid architecture is pretty high everywhere.

Secondly, knowledge of grid computing has reached a point where users are ready to differentiate between the grid types that can be utilised - from constrained cluster grids used for running a single application or set of services, to enterprise grids and shared community grids. This compares to the general response of 'Grid? What's that?' we received when we first interviewed people about the technology nearly three years ago.

This increase in awareness and knowledge is leading to faster grid adoption than we expected. We initially envisaged grids becoming mainstream between 2008 and 2012; we already have 20 per cent of respondents piloting or utilising grids in one form or another, and this indicates that real mainstream grid usage will be around 2007 to 2010. The focus of High Performance Computing (HPC) grids (those for scientific mathematical modelling and research use) is beginning to disappear - users are looking at how grids can be used in more commercial environments, and in ironing out the peaks and troughs in their day-to-day system loads.

The drive behind this move to grid has been the realisation that adoption is not as much of a revolutionary move as first feared. Grid is dependent on virtualisation of the environment - of the network, storage, databases, hardware and services - and this is happening due to the move towards open standards and the acceptance of web services and service oriented architectures (SOAs).

In most cases, no new hardware is required - the grid can absorb existing hardware and utilise the same network, storage and database components.

Historically, applications have needed to be reworked to enable them to make the most out of a grid but new additions to middleware platforms (such as improvements to Oracle's Application Server platform and IBM's latest version of its WebSphere Application Server) provide a platform that 'spoofs' the application into being somewhat grid-aware. As management tools mature to better manage a grid environment - and to provision and de-provision the required stack of operating system, application server and services logic - the grid infrastructure becomes more of a realistic option.

The vendor community is now beginning to push these messages, rather than just feeding the perception of lower costs and higher utilisations. As vendors push their service-based offerings, the users find themselves with a highly standardised and virtualised infrastructure - ideal for a move to a grid architecture.

So - what problems remain for users interested in grid computing? The perennial problem of security does not seem to be a show-stopper. The majority of respondents to our research stated that although security is an issue, it is something that can be reasonably easily dealt with. Cost is perceived similarly - not just the cost of implementing a grid infrastructure but of gaining corporate support and identifying and allocating costs for such a decentralised project.

The major issue for the respondents was getting the broad buy-in to an enterprise grid approach. Thus, many are looking more towards implementing constrained, departmental grids as a starting point. Once these grids are in place, the idea will then be to join them up. But the politics of business are still seen as an issue, with departments arguing over who should be able to utilise 'their' computing power - and how much cross-charging there should be.

Quocirca believes the vendor community needs to up its game - licensing is a thorny issue in the new service-oriented world, where per cpu/per user licences are being frowned upon by the users. However, users are unsure of how they want to pay for software in a grid environment; our research shows a split vote as to whether fixed or variable pricing is the way forward.

The general view is that grid is going to happen. For those who have a high legacy content to their existing environment, this is seen as problematic but services from the likes of Attachmate, IBM, Neon and Seagull help to pull mainframe systems into the services environment, while systems such as HP's Superdome, IBM's iSeries and high-end Solaris boxes are relatively easy to pull into the mix due to the usage of open standards on these platforms. Grid management software is appearing - not only from the high-end players such as DataSynapse and Platform but also from the main systems management players like BMC, CA Unicenter and IBM Tivoli.

As virtualisation is accepted into the heart of the infrastructure, grid becomes more of a possibility. With companies looking at overall asset utilisation of less than 15 per cent on average, it becomes an easy decision to move to solutions that can up this to greater than 40 per cent without much redesign being required. That the infrastructure then enables greater business flexibility makes this decision even easier - to a level where our US brethren would call it a 'no-brainer'.

But there remains one issue - the existing status of IT within an organisation and how much money is available to invest in new functionality.

During this research, Quocirca took the opportunity to ask users how involved the business was with the decisions being made in the IT world - and got some very interesting answers.

We are all aware of how today's IT budgets are predominantly spent on fire-fighting, or maintaining the existing environment. Received wisdom is that 75 to 80 per cent of all IT budgets are spent in this way. For those companies that do not involve the business in the IT decision-making process, this split is correct. However, for those companies where the business is heavily involved in IT decisions, we see that 42 per cent of these companies spend more than 75 per cent of their IT budgets on new investments. And these are the companies that are leading the way with grid implementations.