Intelligent Overload Control for Composite Web Services

Abstract

Abstract.In this paper, we analyze overload control for composite web services in service oriented architectures by an orchestrating broker, and propose two practical access control rules which effectively mitigate the effects of severe overloads at some web services in the composite service. These two rules aim to keep overall web service performance (in terms of end-to-end response time) and availability at agreed quality of service levels. We present the theoretical background and design of these access control rules as well as performance evaluation results obtained by both simulation and experiments. We show that our access control rules significantly improve performance and availability of composite web services.

Integrating Fault-Tolerance Method into Service-Oriented Computing

“…Thus, each invocation model may have a different fault likelihood value (even for the same service invoked). Phase I is divided into multiple steps ( Figure 1): calculating the service's weight in the composition relative to all services invoked (λ) (similar to the approach used in [13]). We use this value of λ as one of the basic constructs, calculating the time a service takes to execute in relation to the total execution time remaining in the composition (λ ), calculating the internal history value (Δ i ) using a Hidden Markov Model, and calculating the external history value (Δ i ) using clustering and reputation.…”

Automated self-healing framework for service-oriented systems

“…Thus, each invocation model may have a different fault likelihood value. Phase I is divided into multiple steps: calculating the service's weight in the composition relative to all services invoked (λ)(similar to the approach used in [14]). We use this value of λ as one of the basic constructs, calculating the time a service takes to execute in relation to the total execution time remaining in the composition (λ ′ ), calculating the internal history value (∆ i ) using a Hidden Markov Model, and calculating the external history value (∆ ′ i ) using clustering and reputation, the interested reader is referred to our prior work in [16], [11].…”

Assessing Fault Occurrence Likelihood for Service-Oriented Systems

“…As mentioned earlier, λ is the ratio of the time that is needed to complete the service execution, divided by the total time of completing the execution of the whole system. Similar to the approach used in [10], we use this value of λ as one of the basic constructs Fig. 3.…”