In the last two newsletters we discussed some of the impact that virtualization is having on application delivery. In the next two newsletters we will discuss the impact that cloud computing is having on application delivery. With that in mind, this newsletter will discuss some of the issues associated with using a public cloud computing solution to support the delivery of business critical applications.
One of the top management and optimizing challenges facing IT organizations over the next year is to ensure the acceptable performance of a small set of business critical applications. However, the service-level agreements (SLA) that are associated with public cloud computing services such as Salesforce.com or Amazon’s Simple Storage System are generally weak both in terms of the goals that they set and the remedies they provide when those goals are not met. Hence, it will generally not be possible for a company to use a public cloud provider to support business critical services. In addition, organizations that use public cloud computing services do so with the implicit understanding that if the level of service they experience is not sufficient for their needs, their primary recourse is to change providers.
Just comparing the SLAs offered by cloud computing service providers is tricky in part because there is not a standard way to create an SLA. For example, consider the availability of an Infrastructure as a Service (IaaS) solution. On the surface, availability appears to be a well-understood concept. In fact, vendors often have differing definitions of what constitutes an outage and hence, what constitutes availability. For example, within Amazon’s EC2 offering an outage is considered to have occurred only when an instance is off line for 5 minutes and a replacement instance cannot be launched from another Availability Zone within Amazon’s geographical region. Not all IaaS providers have a similar definition of availability.
Insight into the availability of a number of IaaS solutions was provided by Cedexis at the Interop conference in May, 2011 (Comparing Public Clouds: The State of On-Demand Performance, Marty Kagan, President and Co-Founder, Cedexis). Cedexis presented data that represented roughly 17 billion measurements that were taken between March 15, 2011 and April 15 2011. As shown in Figure 8, none of the IaaS providers that were monitored delivered availability that was greater than 95%,
In the next newsletter we will continue the discussion of how cloud computing impacts application and service delivery. In the mean time, we encourage you to download for free the first couple of serialized publications of the 2011 application and service delivery handbook. They can be found here.
