Dynamic Workload Peak Detection For Slack Management

Aleksandar Milutinovic1,  Kees Goossens2,  Gerard J.M. Smit3
1Mr, 2Prof. Dr., 3Prof. Dr. ing


Abstract

In this paper an analytical study on dynamism and possibilities on slack exploitation by dynamic power management is presented. We introduce a specific workload decomposition method for work required for (streaming) application processing data tokens (e.g. video frames) with work behaviour patterns as a mix of periodic and aperiodic patterns. It offers efficient and computationally lite method for speculation on considerable work variations and its exploitation in energy saving techniques. It is used by a dynamic power management policy which has low overhead and reduces both requirements for buffering space, and deadline misses (increase QoS).

We propose a speculative policy with peak and phase detector which utilizes and exploits patterns in workload for the power saving purposes and reduction of buffer space. Dynamic detection of periodic and aperiodic mode is applied. Also, dynamic detection of peaks’ amplitude and their variable inter-arrival distance is detected. Based on this, we apply adaptive power management by reducing operating voltage and frequency. On top of that, we pro-actively generate slack as a safe slack margin to reduce number of deadline misses that might happen due to this speculative and non-conservative approach.

We evaluate our policy in experiments on MPEG4 decoding of several different input sequences and present results.