Service Level Agreements (SLAs) in the Cloud are going to change radically over the next few years, so much so that they could become a legal quagmire for the unsuspecting and unprepared.
This is just one of the issues taking centre stage in the mind of Dr John Manley, director of HP’s Automated Infrastructure Laboratory in Bristol. He was one of the earliest to identify the potential of the cloud and how it would work, and is still one of a coterie of important thinkers on how it will develop.
“Security is an important issue for research,” he said. “It is often over-blown as a problem area but by the same token it is sometime under-blown. One of the key drivers of applied security technologies will be the development and growth of SLAs that are backed by the application of liabilities.”
At one level, he feels, this will mean that SLAs will carry the obvious liability - a financial penalty if the SLA is breached. At the other end of that spectrum it could mean the close linkage of an SLA with a brand name, so that SLA failures damage brand reputation. This will be particularly important to the service aggregators and brokers that he sees developing as an important part of the cloud service delivery marketplace. These will be the customer-facing end of a major re-intermediation of the internet as new businesses emerge to provide specific service functionality and service components.
“SLAs will become customer-facing and be geared to metrics that mean something to them, such as the number of customer transactions per second,” he said. “Federated Service Management across multiple service providers will, therefore, be one of the key capabilities of the aggregators and brokers, as `owners’ of the SLA with the customer.”
The management complexity of this he sees as a key focus. “Getting the proper co-operation of all the service providers participating in delivering the SLA is the goal. So the core technologies – and management tools – for the future will be focused around the issue of the global optimisation of what may be, for the more complex service deployments, many thousands of individual services that go to make up a complete service for a customer.”
He is also sure that one of the key elements of Federated Service Management into the future will be its ability to manage service delivery and SLAs over the long haul. “Many services, once established, will need to run for several years and the management environment will need to cope with service updates, occasional bug fixes and platform updates. This will all need to be accomplished while the service is operating and without breaching a transaction-based SLA,” he said.
This does mean that there is still a need for standards to develop, but they need to appear at the right time rather than too early. We are not bereft of standards as it is but the next standards will be aimed at what is needed in what calls the World Cloud as it develops, and that is what has to emerge.
Management systems will also need to cope with complex workload management based on business metrics. For example, Manley sees the future cloud delivery world being market-driven, where the cost of a machine to any customer is determined by its workload at that moment.
HP Labs has already developed some expertise in this area through running the SE3D programme. This gave a Bristol based community of animators the chance to buy time on the Labs’ resources in order to render their animations. They were given notional amounts of `money’ to buy time in a market environment – the earlier they bought a time slot the lower the price - and if it was found not to be needed they could then sell to another that wanted it, and was willing to pay for it. This, Manley feels, is how cloud service delivery will develop.
“Such processes will need to be fully automated as the demand fluctuations, SLA implementation, contracts and federated service setup will happen so fast no team of humans could manage it,” he said.
Manley sees the cloud dividing into two types of basic service. One is the Horizontal Cloud – the generic services required by all. The other is the Vertical Cloud, where the need is to bring together the elements specific to the needs of a vertical marketplace. This can be considered as PaaS +.
With the Horizontal Cloud the challenge is creating partitions that meet the need to isolate the services that are used by a particular user. This is leading to the development of Cells, self-contained, complete service packages that are built from the generic building block services.
Cells are one Manley’s current prime targets. “These will need to be secure, isolated partitions within a virtual infrastructure,” he said. “They will be able to scale up and down and will be model-based and totally de-centralised. Users will be able to submit a template setting out what is required and the system will be fully autonomic. It will, for example, be able to modify its own template to scale up or down, according to the workload put on it.”
The modifications allowed will then be the subject of policies. “Policy is set to be the new `big thing’,” he said.
Post new Comment