Most European companies are flying blind because they have no insight into the level of load or stress on their server infrastructures, recently released research has warned.
According to the study by QNB Intelligence, those companies that do have some insight into network loading levels indicate that one third of servers are running at or close to capacity; but overall only about 60 per cent of average server capacity is be used.
QNB’s parent company, analyst firm Quocirca, believes that grid computing has the potential to address some of these problems, but adds the caveat that further work needs to be done to improve security and privacy in relatively immature commercial grid offerings.
"The survey shows that nearly half of medium and large European companies have considered grid computing in some form or other, such as computing on demand or utility computing," said Clive Longbottom, head of research for Quocirca.
"Cost is seen as the top issue with grid, which conflicts with the understanding that grid brings server consolidation benefits. Interestingly, threat to Intellectual Property comes next as a perceived issue, which puts the onus upon vendors to ensure rigorous attention to security and privacy in their implementations of grid computing."
Longbottom believes that the biggest problems holding back grid computing centre on ill-defined standards and major issues in provisioning the rapid layering of functional components onto the grid. This provisioning, Longbottom said, is vital to enable a heavily loaded process to run in a more parallel way, but not to be dependent on specific compute assets within the grid.
“The Global Grid Forum (GGF) has not yet managed to get an agreement on provisioning standards and so IBM, Oracle et al do it all in different ways. If this continues, when two grids meet (which they will at some point), each one will try to provision the same compute assets - and the whole thing falls apart,” Longbottom warned.
There are also serious technological issues associated with grid management tools that the Quocirca identifies as not yet having been adequately addressed: “It's fine being able to take the approach of looking just at the compute, network and storage assets and maintaining a view of what's working or not - the problem comes with the scheduling and marshalling of functional assets (the bits that are trying to fulfil business processes).
“But we need to be able to split the process down to a set of tasks, and get each of these tasks to be fulfilled across the grid. The problem here is that the data coming back has to be pieced back together - and this needs a grid controller function. But this is an interim problem - most of the big management vendors are working on the problem.”
The August and September 2003 survey of 307 respondents across eight countries and 10 industries, commissioned by database vendor Oracle, reveals a classic early adoption attitude to grid computing, with almost half of respondents indicating that they have considered the technology.
The top two areas for consideration of grid computing were found to be enterprise applications (such as enterprise resource planning and customer relationship management) and databases. However, 45 per cent of companies said they feel that grid computing will not become a mainstream commercial reality for three years or more.
Quocirca’s report concludes that “the usual suspects” - IBM, Oracle and HP – are likely to become the dominant players if grid technology hits the mainstream. Microsoft, however, is expected to go after a chuck of the grid pie through .NET - but will not deliver actual product, just base capability.