One of the interesting side effects of the rapid adoption of Cloud Computing by the enterprise is the impact this adoption will have on the design and delivery of IT service processes.
In his article Assessing cloud providers, Frank Ohlhorst reminds us that “moving to the cloud is primarily a business decision” dependent on the metrics of ROI (Return On Investment), performance, sustainability and suitability to task.
Managers, writes Ohlhorst, must be prepared to do the following:
– audit the target applications and business processes impacted to create a cost-benefit-risk analysis that compares a traditional client/server solution to a cloud-based solution.
– audit the cloud services provider, including an assessment of geographic redundancy, packet transport performance, latency and service guarantees.
– audit the business’s own ISPs, including performance at connecting points, failover capabilities and guaranteed throughput rates to and from the cloud services provider.
– monitor and frequently evaluate service and performance elements.
Thus, Ohlhorst tells us, “one of the first steps for choosing a cloud service provider is to evaluate the level of service offered and the guarantees behind that service.” His view is that the Service Level Agreements (SLAs) must be scrutinized under three specific lenses: data protection, continuity and costs.
While this is a traditional IT view, and seems quite logical, we disagree with his suggestion that IT Managers can turn to the Keynote Internet Testing Environment (KITE) and Internet Health Report to measure performance.
Why? Because these are uptime measures, not measures of service performance.
If you’re familiar with ITIL V.3, you’ll recognize this service model overview:
In the ITIL world, service management can be broken into the following components:
– Service Strategy
– Service Design
– Service Transition
– Service Operation
– Continual Service Improvement
Traditional IT systems management thinking leads us to associate systems availability with service availability, so that if a network component is running normally, we assume that the services running across that network component are also running normally.
This is largely the view being taken by the traditional systems management companies. It is what we are seeing in announcements like this one from BMC Software and Amazon.com.
But the cloud service model is different, and – while it’s great to see BMC extending its enterprise systems management platforms to incorporate Cloud infrastructure – Cloud computing brings about a different measure for service performance, best exemplified by a new breed of cloud computing management vendors like Nimsoft. Their view is as follows:
The “pay-as-you-go” nature of cloud computing breaks the link between component and service performance: typically, organizations pay for capacity or throughput, rather than specific components. Plus, the highly dynamic nature of the computing infrastructure that exists in the cloud makes traditional CMDB (or simple list) based systems management virtually impossible to implement. All the traditional server and network reporting that shows 99.999 up-time will become secondary and probably irrelevant for future service level management and reporting. What this means is that synthetic transaction monitoring–that is, generating, monitoring, and reporting on simulated service requests–will be of paramount importance.
This perspective puts an interesting twist on ITIL’s IT Service Management model. Since there is no way to predict which cloud computing infrastructure components are accessible at any point in time, service delivery processes in the enterprise – and SLAs from cloud computing service providers – need to be all about service reliability rather than component reliability. This is a paradigm shift.
As we have written previously, cloud computing is unleashing the potential of SOA (Service Oriented Architecture) applications. In a world of SOA applications running on Cloud infrastructure, the concepts of IT service delivery in the enterprise and SLAs from service providers will rest upon services and processes that can run on any infrastructure components within the cloud. The notion of using discrete infrastructure components as the basis for measuring service quality goes away. This is the philosophy of the new breed of cloud systems management providers: the focus of availability and performance measurement moves toward measuring the user experience.
And, as this transition comes about, what happens to CMDB-based systems management? How do we think about the CMDB when the management of these infrastructure parts is abstracted even further away from application peformance? Does anyone see a new “cloud edition” of ITIL service delivery on the horizon?
Once again, there is an opportunity here for service providers to seize the initiative.