Automation of validation environment for ST Microelectronics Set Top Box Audio Firmware

Abstract

Audio Firmware of Set Top Box needs to go through regression testing and validation process as per the certi cation criteria of rmware whenever any audio codec of rmware is upgraded. Since the amount of work needed to validate these audio components is quite high, there is a need for a uni ed validation environment that automates the certi cation process and provides detailed reporting of failure and certi able cases. When cluster of workstations is available for validation process there is a need of global scheduler which helps to make e ective scheduling decision by predicting resource availability in cluster. By analyzing the historical CPU load data of workstations a prediction of average CPU load of workstation can be estimated. Validation process can be hastened by assigning jobs to available resources . Various linear and non-linear time-series prediction methods are developed for CPU load prediction where CPU usage data is treated as a time-series. This thesis devises a CPU load forecast model to predict CPU load from analyzing historical CPU load data using su x tree. A probability based one-step ahead prediction table is prepared according to frequently occurring patterns in time-series data. This probability based prediction approach is capable for one-step and multi-step ahead prediction with improving prediction table according to prediction hit-miss. Proposed method has lower time-complexity, better indexing power compared to other linear and non-linear time-series forecasting methods.