AESOP home


Sliding Hidden Markov Model for Evaluating Discrete Data

Tiberiu Chis

Conference or Workshop Paper
10th European Workshop on Performance Engineering (EPEW 2013)
August, 2013
Lecture Notes in Computer Science
Volume 8168
Springer Berlin / Heidelberg
DOI 10.1007/978-3-642-40725-3_19

The possibility of handling infrequent, higher density, additional loads, used mainly for on-line characterization of workloads, is considered. This is achieved through a sliding version of a hidden Markov model (HMM). Essentially, a sliding HMM keeps track of processes that change with time (i.e. changing the observation set at different stages of analysis) and the constant size of the observation set helps reduce the space and time complexity of the Baum-Welch algorithm, which now need only deal with the new observations. Practically, an approximate Baum-Welch algorithm, which is incremental and partly based on the simple moving average (SMA) technique, is obtained, where new data points are added to an input trace without re-calculating model parameters, whilst simultaneously discarding any outdated observation points. The success of this technique could cut processing times significantly, making HMMs more efficient and thence synthetic workloads computationally more cost effective. The performance of our sliding HMM is validated in terms of the means and standard deviations of observations (e.g. numbers of operations of certain types) taken from the original and synthetic traces.

PDF of full publication (94.8 kilobytes)
(need help viewing PDF files?)

Information from