Rapid and accessible performance evaluation of complex software systems requires two critical features: the ability to specify useful performance metrics easily and the capability to analyze massively distributed architectures, without recourse to large compute clusters. We present the unified stochastic probe, a performance specification mechanism for process algebra models that combines many existing ideas: state and action-based activation, location-based specification, many-probe specification, and immediate signaling. These features, between them, allow the precise and compositional construction of complex performance measurements. The paper shows how a subset of the stochastic probe language can be used to specify common response-time measures in massive process algebra models. The second contribution of the paper is to show how these response-time measures can be analyzed using so-called fluid techniques to produce rapid results. In doing this, we extend the fluid approach to incorporate immediate activities and a new type of response-time measure. Finally, we calculate various response-time measurements on a complex distributed wireless network of O(10^{129}) states in size.

This tutorial surveys the fundamental results of the theory of martingales from the perspective of the performance engineer. We will present the fundamental results and illustrate their power through simple and elegant proofs of important and well-known results in performance analysis. The remainder of the tutorial will introduce the martingale functional central limit theorem and semi-martingale decomposition methodology for the characterization and proof of heavy-traffic limit results for Markovian queueing systems.

Jeremy T. Bradley, Nigel Thomas, Richard A. Hayden, Anton Stefanek; Invited Response to Computer Journal Lecture by Prof. Jane Hillston, The Computer Journal, V

Capturing energy consumption directly from a stochastic behavioural model is a computationally expensive process. Using a so-called fluid analysis technique we are able to access accumulated reward measures in much larger scale stochastic systems than has been previously possible.These accumulated rewards are ideal for deriving energy and power consumption from stochastic process models. In previous work, it has been shown how to derive a set of ordinary differential equations (ODEs) whose solutions approximate the moments of component counts in a continuous-time Markov chain(CTMC) described in a stochastic process algebra. In this paper, we show how to extend the method to provide rapid access to moments of accumulated rewards in CTMCs. In addition to measuring the amount of energy used by a system, we are also interested in the time taken to reach a particular level of energy consumption. In reward terms, this is a so-called completion time. In this paper, we are able to use higher moments of rewards to give us access to completion time distributions.

We introduce a low-level performance modelling formalism, Shared Transaction Markov Chains (STMCs), specifically designed for the capture and analysis of massively parallel stochastic systems through fluid techniques. We introduce the notion of a shared transaction between concurrently running Markov chains which allows a multi-phase synchronisation to accurately represent complex cooperation between modelling components in a compositional manner. We demonstrate the new modelling formalism on four distinct models and show how fluid analysis may be performed, with results, where appropriate. Our contribution is that this is the first such system tailored to the fluid performance analysis of transaction-based systems as found in computing applications such as peer-to-peer networks, Web architectures and Publish-Subscribe networks. The second contribution is that STMCs permit composed phase-type distributed synchronisation which is more useful from a transaction modelling perspective.

We present generic extensions to the GCspy visualisation frame-work that make it suitable for tracking the way continuous dynamic memory allocators such as dlmalloc or incremental and concurrent garbage collectors make use of heap memory. These extensions include sample-driven client-server communication, in-cremental stream updates and client-controlled stream update frequency. Additional extensions to the current GCspy client are also described. These include hierarchical driver grouping and hierarchical visualisation, zooming, and the ability to define and view relationships between tiles in different spaces. We also introduce a heuristics engine that is responsible for flipping GCspy from its decoupled 'observation' mode to a synchronous 'single-step' mode, and describe a backtrace facility that can trace the server-side call sequence that led to the triggering of a specified event, such as the allocation or freeing of a block of memory. This enables aspects of the allocator (fragmentation, block ordering, splitting and coalescing policies, etc.) to be understood in the context of a particular application and potential optimisations to be identified. The effectiveness of the enhanced framework is demonstrated with a complete integration with dlmalloc. The framework is evaluated in terms of both performance and its ability to explore contrived modifications to dlmalloc's coalescing policy.