Efficient descriptor-vector multiplications in stochastic automata networks
Journal of the ACM (JACM)
Using the exact state space of a Markov model to compute approximate stationary measures
Proceedings of the 2000 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
Structured analysis approaches for large Markov chains
Applied Numerical Mathematics
Structured Solution of Asynchronously Communicating Stochastic Modules
IEEE Transactions on Software Engineering
Performance Evaluation of Computer and Communication Systems, Joint Tutorial Papers of Performance '93 and Sigmetrics '93
Performance Evaluation of Complex Systems: Techniques and Tools, Performance 2002, Tutorial Lectures
A Data Structure for the Efficient Kronecker Solution of GSPNs
PNPM '99 Proceedings of the The 8th International Workshop on Petri Nets and Performance Models
Adaptive decomposition and approximation for the analysis of stochastic petri nets
Performance Evaluation - Dependable systems and networks-performance and dependability symposium (DSN-PDS) 2002: Selected papers
Memory Efficient Calculation of Path Probabilities in Large Structured Markov Chains
QEST '08 Proceedings of the 2008 Fifth International Conference on Quantitative Evaluation of Systems
Hi-index | 0.00 |
We consider a class of Markov chains modelling concurrent activities with phase-type (PH) distributed duration. Specifically we focus on systems whose dynamics is characterised by the parallel execution of a number of active jobs and such that the termination of an active job causes the remaining active jobs to be restarted from scratch (preemptive restart policy) or deactivated and can lead to the activation of new jobs. In such a framework, the state-space can be partitioned into disjoint sets (called macrostates). Each macrostate represents the parallel execution of the (currently) active jobs and a transition to another macrostate corresponds to termination of one of the active job. Models of this kind are subject to the phenomenon of state space explosion, to an extent which is proportional both to the level of concurrency (i.e., the number of active jobs in a macrostate) and to the dimension of the phase-type timing (i.e., the number of phases used to model jobs' execution). Although techniques exist that allow for handling the infinitesimal generator matrix of such models in a memory efficient manner (e.g., in terms of Kronecker expressions), classical Markovian analysis remains hindered by the memory requirements for storing the vector of the steady-state or transient probabilities. In this paper we show that a Markov chain satisfying the above assumptions can be analysed by calculations performed on the (small) matrices describing the durations of the jobs without the explicit storage of the (large) vectors of transient of steady state probabilities.