With the concept of utility computing in such a stage of infancy, parties considering embarking on such a drastic sea change should make adequate provision in the contract to protect all on board.

Flexibility and Adaptability

The key to the utility model is adaptability.  When the needs of the user grow, the service provider must respond appropriately, controlling resources to fulfil requirements.  In order to make the project financially viable, customers looking to take advantage of utility computing services will, by definition, have volatile IT needs and an expectation that use will fluctuate dramatically throughout the contract term.  This means that the description of the services being supplied must be defined with precision, since it will include both current technologies and future technologies.

When the breadth of the IT services required is constantly changing, inevitably, the scale upon which performance is monitored changes too.  Unlike a standard outsourcing agreement, utility computing demands that Service Level Agreements (SLA’s) be as adaptable as the service against which they are measured.  However fixing strict benchmarks for performance at the outset may lead to practically unachievable targets as the needs of the company evolve.  As such, customers may prefer a tiered approach to SLA’s, or include other mechanisms to ensure that expectations of both the customer and the supplier remain realistic, goals achievable and deliveries match the fees paid.

The same is true of service credits, the financial “penalty” imposed on a service provider for poor performance. What was an appropriate incentive to perform when IT demands were small will probably be insufficient should the scale of operations alter by any sizeable increment.  Provisions in the agreement should address how the quanta of service credits payable should change in accordance with the extent of IT operations.

Handling Data

As with any outsourced services, especially where essential data handling is performed in a location remote from the customer’s premises, the issues of data security arise. The customer is no longer able to physically monitor the operations of the service provider, has no control over their personnel or hardware and entrusts safe and effective management solely to the provider by virtue of the role delegation.

By definition, the technical infrastructure relied upon to provide utility computing only exacerbates the problem.  As the demands of the customer change, so too the utilization of the hardware must respond, compounding the possibility for errors.

Structuring both servers and storage devices within a network to operate as one adaptable resource requires both effective monitoring and control.  Oracle, for example, states that users will be able to smooth out peaks in demand by spreading the workload across a grid of servers.  Once the peak has passed, the grid can be reconfigured to run different applications.  Accordingly, in theory there should be higher availability since there is less likely to be a single point of failure.  Utility computing not only demands that servers be able to swap roles on demand, have their hard drives completely wiped but may even need to change operating systems in their entirety.  To cope with the task, service providers modify existing systems tools to perform this duty.  For IBM, The Utility Management Infrastructure, code-named Blue Typhoon, performs at the core of their resource, interpreting needs and managing resources.  This key player, essentially binding the network together as a single responsive unit, is also the key weakness.

With adaptability comes increased risk in performance, security and data integrity.  We have to ask, who will bear the added cost of these risks?  Will the increased price of evolutionary technology be factored together with higher premiums to protect against problems?

Another key difference between a standard outsourcing arrangement and a utility computing model will be how the customer pays for the use of software.  Currently most of the software vendors operate on models whereby they 'sell' licenses for each product.  The licenses concerned may function in a variety of fashions, but most often operate on a 'per server', 'per CPU' or 'per user' basis.  Some applications may function on a 'concurrent' basis whereby the license covers simultaneous use by a specified number of users rather than of individually identified users.

It is apparent that in order to function cost effectively, utility computing requires that software licenses be made available quickly, and that consumers only incur the license fee while the software is actually in use.

As with any large outsourcing contract, workability is the key to success. Although termination is always an option, better results may be had by providing incentives for success through the contractual provisions.  Fluctuation in demand for services must therefore be reflected in service monitoring mechanisms.  Add to this the increased risks involved with networking resources for delivery of IT services as an on-demand utility, and the greater need to apportion risks and to provide for potential pitfalls.  To what extent these factors affect the price tag, and which party foots the bill are fundamental in negotiating the final contract.

Alan Steele-Nicholson is head of the IT-Telecom law department in the Rotterdam, the Netherlands, office of the global law firm Simmons & Simmons, and can be contacted on alan.steelenicholson@simmons-simmons.com. Peter Brudenall is Senior Lawyer in Simmons & Simmons’ London office, and can be contacted on peter.brudenall@simmons-simmons.com. Both are specialists in IT and telecom outsourcing law.