Performance management in cloud computing
Have you experienced the frustration of waiting for information to load on a computer screen when your IT systems were running slowly? Now multiply those emotions and their impact on your productivity by the number of people in your organization, and you start to appreciate the overall business impact of poorly performing applications.
Consider a company that shares the responsibility of delivering applications with a service provider while still being accountable to its end business users. Sharing this level of responsibility requires a greater degree of trust in that service provider than a traditional service model, especially with regard to data integrity and confidentiality, service availability and performance.
Transparency is essential when sharing the responsibility of delivering IT services to end users, explains Richard Fisher, Managing Consultant, Orange Business Services. The IT department and service provider must be completely transparent in managing the different components of the infrastructure that deliver the IT service. This will eliminate finger-pointing and provide the best possible application performance.
Identifying and solving problems
The first challenge for cloud application performance is detecting and solving problems within agreed timeframes. Operational support groups are typically compartmentalized and specialize on each component of the application delivery chain. Cloud computing adds to the complexity by introducing an external organization. The organizational separation between support groups, combined with the added complexity of virtualized technology, makes it difficult to identify the root cause of application performance issues.
In addition, the proliferation of consumerization means that countless variants of end-user devices, operating systems and browsers will access cloud based applications. As a consequence, applications are no longer certified for use on client devices, but rather problems will be handled dynamically and reactively. This will require the means to detect, quickly isolate and find the root cause of performance problems at a much faster rate than achieved currently.
Managing expectations and perceptions
An end user could wait months before reporting a performance-related problem, unless it means that he cannot access a business-critical application that he needs to do his job. During the brownouts, he won t be as productive as he would otherwise be and will probably share his dissatisfaction with his colleagues. This negative build up around a particular application will most likely escalate if it is delivered by an external service provider.
End-user experience monitoring is key to cloud computing, explains Fisher. To avoid widespread application performance problems you need to spot problems before you are told about them.
Delivering SLAs that count
As the market matures and cloud computing offers become more mainstream, it is expected that the most successful providers will be those who have built a partnership with their customers – and don t just sell CPU minutes, storage and bandwidth.
Cloud service providers need to develop a partnership based on trust. They need to speak the same language as the enterprise customer and understand the business requirements of the service. Traditional SLAs based on technical KPIs for round-trip time and CPU minutes do not map directly to business requirements.
In most cases, it doesn t matter how well the service provider delivers against its own targets because the customer may find it difficult to relate those targets to their internal KPIs. The service provider needs to measure the performance of the cloud service in terms that directly map to the requirements of the business users.
Traditional network or application SLAs such as latency, loss jitter, bandwidth, mean time to repair are insufficient, explains Fisher. What we really need to offer are services defined by end-to-end SLAs based on how the end-user experiences the application or IT service. These can be measured in terms of quality metrics for availability, response time, reliability and service-recovery-time.
Service providers need to design KPIs that map to business performance metrics and offer these as SLAs. Fundamental to this are the end-user-experience KPIs, the measurement of which only the service provider can support, as it requires visibility of the end-to-end system through monitoring of the network traffic.