Scheduling soft real-time jobs over dual non-real-time servers

Multiple non-real-time servers can be used to exploit the redundancy of soft real-time systems in improving the system's ability to meet response time goals. The scheduling policies commonly used are the Balance strategy (even distribution of loads) and the Chop strategy (load partitioning according to job slackness). A simple analytical model is used to analyze the effectiveness of the two schemes. Results showed that the Chop strategy significantly improves the performance of real-time systems, especially in moderate loads and wide variations in job slackness.