What are service levels and service credits:

Service levels allow parties to measure in an objective manner the performance of the service. Service credits aim at ensuring, through financial incentives, that the services meet the agreed service levels. In a software-licensing arrangement the customer can make a technical assessment of the software to be provided and decide whether or not the software meets the customer's needs. By contrast, in a cloud-computing arrangement (as in IT-outsourcing arrangements) the customer is completely reliant on the contractual services descriptions which form the basis of the service to be provided to the customer. Therefore, negotiating and including adequate service level and service-credit arrangements is very particularly important in cloud-computing arrangements. 

Service levels measure the following: (i) service availability; (ii) service response time; (iii) correction time

i. Service availability

Service availability measures the extent to which the cloud-computing service is available to users as a percentage of the time during which the service provider is contracted to provide the service to the customer. In order to determine this a number of issues must be considered, mainly:

·         Point of measurement.

·         Measurement period.

In order to be able to calculate the service availability it is important to agree where the service is measure (or as we said “point of measurement”). Service providers normally aim to establish the point of measurement at their servers, but this argument should not be accepted without question by the customer. Where the transmission is over the internet, there are many different types of internet service provision and the cloud service provider should not be allowed to adopt a potentially low-quality and low-cost approach, with the inevitable impact on service quality, unless the customer understands the approach and has agreed to it.

From the user's perspective, service availability measurement at the user's computer is attractive because the measurement assesses the extent to which the cloud computing service is available to the user. However, this may not be possible if there is no technological method of assessing service availability at the user's computer. A customer needs to have a reasonably sophisticated system infrastructure to make the measurement, but the necessary tools are no longer uncommon.

A service provider can be expected to argue that the service availability measurement at the user computer level is inappropriate as this will introduce downtime (when the service is not available) resulting from the customers' infrastructure failure, rather than from the failure of the cloud service. The service provider may therefore suggest that the point of measurement should be the cloud termination point at the customer's premises.

However, if service availability is measured at this point, it will be more difficult for the customer to assess service availability at an individual-user level rather than at the aggregate level relating to all of the users to whom the service is provided.In addition, the period when service availability is measured needs to be specified, because the choice of the service measurement period has an impact on the calculation of service-level assessment.

While a 24/7 service might appear to be attractive (particularly for global organisations), in practice this can lead to a need for considerable downtime before service credits are incurred. For example, a 98% service level would mean nearly 15 hours of downtime in a 31-day month before it was failed, whereas on an 8.00 am to 6.00 pm weekday service, around four hours of downtime might be sufficient to trigger service credits. Customers should consider the impact that taking either approach would have on their operations. Downtime that does occur out of hours can be more difficult to rectify quickly as engineers are less likely to be available.

With the focus of cloud computing on flexible anytime, anywhere access, out of hours downtime can soon have a detrimental effect on a service that would otherwise have met its service availability targets and which in all other respects is acceptable.In the event the service provider offers a bundled cloud service comprising, for example, e-mail, internet browsing and office applications (such as word processing and presentations applications), the service availability service level should relate not only to the overall cloud computing service availability, but also to the availability of the individual software applications. 

Service response times                                                   

As well as being made available to users, computer systems need to have a prompt response to user inputs. The key measurement of responsiveness is the speed at which the screen responds to user inputs. System responsiveness measures are more difficult to specify in cloud computing-contracts than service-availability service levels. This is partly because the complexity of the request that the service will be measured against can be highly variable.

For example, a complicated data matching and retrieval operation will inevitably take longer than simply inputting data. However, even apparently simple operations, such as printing and saving documents, may be relatively complex in a cloud-computing arrangement and take a long time. Print processing may require activities by the service provider at the server level that may increase the transmission time taken to activate the printer. For example, when an Australia-based user wants to print from a web page provided by a UK-based service provider, the web browser servers in the UK have to be activated in order to transmit a print instruction to the user's printer in Australia.

Equally important from the users' perspective are cloud-computing helpdesk response times. In particular, will the initial response involve providing answers and user support, or will it simply log the answer with a further call back to provide substantive support?

A service provider is likely to use a call-logging system as it enables them to meet the helpdesk response time requirement more easily and the personnel for the phones will be less expensive than where fuller user support is provided. It can be useful to include a specific service level for the number of calls that will be resolved on a first-call basis to avoid a call-logging approach. It is reasonable to expect at least 60% of helpdesk calls to be resolved on a first-call basis. Different measures will be required if support is provided only by e-mail. 

All articles are for general purposes and guidance only and do not constitute legal or professional advice. Copyright 2010 Anassutzi & Co Limited. All rights reserved. Information may be shared or reproduced only if accompanied by the author’s name and bio.