Scheduling Real-Time Components Using Jitter-Constrained Streams.

Abstract

Component-based applications require good middleware support. In particular, business logic should be separated from management code for guaranteeing nonfunctional properties of a system. We present an approach called ContainerManaged Quality Assurance, in which a component container uses nonfunctional specifications of components to determine how to use these components, and which system resources to allocate, to provide certain services with guaranteed nonfunctional properties.As an example, we show how this technique can be applied to automatically allocating CPU and memory resources for components with real-time constraints. To this end, we use a mathematical model based on jitter-constrained streams, a mathematical abstraction of event streams.

Formal specification of non-functional properties of component-based software systems

“…As another example, knowing all components cooperating to provide the service, the container can decide where to place buffers and how to dimension these buffers, thus balancing response time against jitter. This example is discussed in further detail in [33].…”

“…Example 2 (Response time) Given a component with one operation with a worst-case execution time, and information about the stream of incoming requests for this operation in the form of a jitter-constrained stream [32], the strategy described in [33] computes the required number of component instances, the amount of memory required to buffer incoming requests for the operation's service, the task set to be scheduled by the underlying CPU, and the worst-case response time for each operation invocation.…”

“…A typical example of the kind of guarantees given by the system environment is the distribution of request interarrival times. This can be used together with queuing-theory-based techniques to determine the optimal number of components and buffer size to achieve a required response time using components with a known execution time [33].…”