Cloud or not, the fundamental obligations of service delivery don't change. At the end of the day, customers and stakeholders, including business managers and senior executives, really don’t care whether services exist in the cloud or not. They simply expect them to work and perform reliably and the onus is on the provider to prove its worth.
Cloud adoption is typically a decision taken by IT managers, not by business managers, and it’s IT management that is ultimately held accountable for the success or otherwise of a transition to the cloud. Sometimes success or failure is readily visible, but more often than not it is subject to interpretation. For complex IT applications, intermittent service failures, slowdowns and partial failures are common. Unless failures are catastrophic, it may not be obvious whether or not a particular application is meeting a defined service standard.
This leaves the IT manager in the uncomfortable position of debating evidence from support channels or the management hierarchy. This becomes a problem if an application has a dubious service history or there has been a major change in the load, functionality or implementation, such as moving to the cloud.
To assure stakeholders, it’s critical IT managers can demonstrate transitioning to the cloud worked and that applications are available and performing for users. This tends to rule out most conventional forms of technology and application monitoring.
Service levels must be defined
Service levels for applications, and how they are to be measured, must be defined and agreed upon with stakeholders. Surprisingly many large organisations don’t do this correctly and rely on out-of-date, unmeasurable or inappropriate service level definitions. This is a problem for IT managers transitioning to a cloud solution where you may get what you pay for in terms of reliability and performance. Unless service levels are clearly defined IT managers are at the mercy of perceptions of an application’s performance, and will be judged accordingly.
Failures are a part of life
Unfortunately, even well engineered and managed systems sometimes fail. Dealing with these problems quickly is difficult enough even when all of the underlying technology is under your control.
In a cloud environment, one or more third parties control all or part of the delivery and it may not be clear where the problem resides. To minimise system failures IT managers must have a clear, independent view of the service status and performance for each third party in addition to complete visibility of the application. This means being able to quantitatively measure their service performance in a way that can be correlated with the service performance of your applications.
What is service visibility?
So what sort of service visibility should you be seeking?
1. User centric: IT applications deliver services to users, so service visibility must start with the user perspective – exactly as the user sees it. Realistic application usage patterns need to monitored, as well as all the various access modes including web, desktop and mobile.
2. Independent: A service visibility solution needs to be independent of both cloud service providers used and underlying technologies. Technology components may come and go but the obligations to deliver service do not.
3. Comprehensive: It’s important to have accurate visibility of application availability and performance, and be able to relate it to defined service level requirements. For external applications, you need to see Internet service behaviour, not just from within a datacentre. The same method needs to be applied across all applications for consistency, and be adaptable to monitor cloud service providers and end-to-end systems.
4. Repeatable: Any solution needs to repeatable, consistent and robust 24 x 7. It must also be suitable for benchmarking application performance and availability before and after transition to the cloud to gain confidence that the move has not caused problems.
Getting real service visibility
Given that you do need independent means of ensuring your own service delivery, how do you do this?
IT organisations have traditionally deployed a plethora of infrastructure and technology monitoring tools to manage in-house technology components. These tools are mostly useless for providing service assurance to customers and stakeholders; and more often than not rely on deploying agents to the monitored devices.
Yet it may not be feasible or desirable to install monitoring software onto all of the various cloud components. This means traditional infrastructure monitoring approaches are even less for use for service level assurances in cloud models.
End-user experience monitoring will help when done properly
End-user experience monitoring (EUE) does have potential to provide real visibility of the services for which IT managers are accountable. EUE provides a concise view of service availability and performance that is independent of underlying technologies.
It can also provide a good basis for before and after comparison of services as they are migrated to a cloud model to validate that no harm has been done in the process. Without complete end-to-end visibility of application performance and availability you may have cloud but you will have a service shrouded in fog.
Murray Andrews is the chief operating officer at REMASYS