Service Level Agreements (SLAs) have been a staple of the enterprise IT industry for the past couple of decades.
However, in this burgeoning era of cloud computing, the SLA approaches of old are in danger of becoming obsolete.
The real reason for SLAs
SLAs guarantee component, application or solution uptime and penalties are assessed for not meeting these guarantees. While these penalties usually come in the form of service credits, IT clients are not typically interested in service credits. The client simply does not want any unexpected downtime in the first place.
One of the most important aspects of the cloud discussion is choosing a cloud provider that will keep client workloads up and running.The SLA is just a chance for cloud providers to "put their money where their mouth is" and show that their motivations are aligned with their client’s.
Occasionally, SLAs are intentionally crafted by some hosting providers to never actually pay out. They are written so that the sun, the moon and the stars have to align for an outage to meet the criteria required for pay-out.
For example, at least one provider that I’m aware of actually requires clients to buy duplicate resources or extra services to even get a basic SLA on core services. So it is important to understand a hosting provider’s uptime track record and the details of an SLA before an outage occurs and you are left with no recourse.
Rise of the cloud
Cloud computing’s enterprise maturity has arrived at a point in the economic cycle when delivering cost savings and business agility rank near the top of many business priority lists. Businesses can ill afford to jeopardise application availability and performance, as without guaranteed uptime performance and infrastructure flexibility, business revenue can be at risk.
Cloud infrastructures typically include varied IT components such as storage, compute, network, security, etc. The best SLAs typically guarantee application or server accessibility and do not differentiate across each component.
As a result, when considering the shift to cloud, IT managers are now re-evaluating SLA priorities, a trend that moves away from the old school SLA, with its relatively simplistic nature comprising crude homogenous uptime and infrastructure performance measures. The change in business priorities, increased adoption of cloud and today’s difficult economic conditions mean that many companies are heading toward SLAs that are more tailored to their specific application needs
Storm clouds ahead for the traditional SLA model
At present, the majority of SLAs are restricted to basic undertakings from the supplier for a set level of uptime in any given period. This approach ideally suited the pre-cloud era with its low emphasis on application performance where little focus could be placed on the different demands of a business application.
Applications with varying uptime requirements also commonly reside at different companies or within a company.Some applications may be less critical to the business and can tolerate lower uptime in exchange for lower cost. Other applications cannot tolerate any amount of downtime without resulting in a significant impact on the business. This is essentially where the inflexibility of the current standard of SLA is exposed.
As an example, some companies have internal back office applications, development environments and other workloads that can tolerate some downtime without greatly impacting the business, whereas client-facing e-commerce websites, financial transactions and other critical workloads require continuous uptime.
Using multi-tiered SLAs allows the selection of the appropriate level of service to closely match the criticality, cost and performance requirements of each application. Some more advanced cloud providers enable clients to build different performing IT environments with SLAs, performance and cost points matching their varying application needs from within the same cloud infrastructure.
This cloud’s silver lining
When evaluating cloud services, businesses must consider the SLAs available to them and ensure that they extend far enough to support the most critical business processes, while complying with company best practices. Basic availability in the cloud era should be taken as standard, which consequently minimises its importance in this new type of SLA.
This makes an ideal SLA one that is flexible and robust enough to afford the business greater control over the costs of ensuring the appropriate level of performance for each application. Paying for the same SLA from a test app deployment to a mission critical process no longer makes sense, and erodes the cost benefits of using cloud infrastructure.
Companies migrating to cloud hosting should not have to compromise on performance or the ability to fine tune their infrastructure. The onus should be on the cloud provider to supply systems and SLAs that comfortably fit with the client’s IT infrastructure needs instead of forcing the client to alter their architecture and cost base to fit a SLA.
Posted by Reed Smith, director of hosting product management, Savvis