Monitoring as an SLA-oriented consumable service for SaaS assurance: A prototype

Whether an enterprise is deploying their applications to virtual machines or ´real´ machines, monitoring of those applications is often critical in meeting key business priorities. The monitoring domain is a mature and competitive one with many offerings ranging from traditional closed proprietary products to open-source collaborative projects. Whilst the offerings have continued to increase in their core monitoring capabilities and sophistication, one area which has received less attention is the programmatic control of monitoring configuration; most monitoring solutions provide little to support the solution designer´s goal of loose coupling with high cohesion between the service management layer and the monitoring layer. There is still less support for such a Service Management Layer to specify it´s monitoring requirements in terms of the SLA the services it manages are being offered against. The reality is that monitoring is often configured manually on an ad-hoc basis, resulting in an evolved monitoring configuration which is poorly structured, inefficient and unmanageable. This paper presents the authors´ work in progressing a design which allows a service management layer to consume monitoring as a service programmatically as per the Service Oriented Architecture paradigm. This makes monitoring auditable, flexible and efficient. Further, the design shifts the primitives used to define the monitoring requirements away from machines, ports and intra-machine services (e.g. HTTP) to the SLA and the distributed services which a whole set of diverse machines co-operate in order to support (e.g. an instance of a SaaS CRM enterprise app). The aim is to allow the operator to promote the SLA such that their systems can optimise customer satisfaction while minimising their cost base. Whilst this approach offers benefits for the management of all data centres, those which are retailing a catalogue of quality-assured, cloud-based SaaS offerings may see the greatest g- - ains as the design elegantly allows for instances of the offerings to be deployed quickly and scaled automatically, including all the required SLA-driven monitoring configuration.