Towards Budgeting in Real-Time Calculus: Deferrable Servers

Abstract

Abstract. Budgeting of resources is an often used solution for guaranteeing performance of lower priority tasks. In this paper, we take a formal approach to the modeling of a deferrable server budgeting strategy, using real-time calculus. We prove a scheduling theorem for deferrable servers, and as a corollary show that an earlier claim of Davis and Burns, that periodic servers dominate deferrable servers with respect to schedulability, no longer holds when the context of the comparison is slightly generalized.

Response Time Analysis of Hierarchical Scheduling: The Synchronized Deferrable Servers Approach

“…Significant research has been done on hierarchical scheduling [14,16,17,28,30]. On uniprocessors, RTA of preemptive tasks executed by fixed-priority bandwidthpreserving servers, e.g., DS, has been studied [14,16].…”

“…On uniprocessors, RTA of preemptive tasks executed by fixed-priority bandwidthpreserving servers, e.g., DS, has been studied [14,16]. These approaches are based on the fact that the job with the maximum response time is released at a critical instant.…”

“…Consider two tasks: τ 1 = (100, 18, 100) and τ 2 = (150, 34, 150) processed by a 2-SDS = (20,14,10). Take as time 0 the beginning of the replenishment period within which a job J 2 of τ 2 is released.…”

“…In Figure 1 (c), if a job J 3 is also released at time 10 together with J 1 and J 2 , R 2 = 60 is longer than R 2 = 54 in Figure 1 (b). Example 4.2 indicates a significant difference between the RTA for hierarchical scheduling and the RTA for dedicated scheduling [13,14,16,20] where only higher-priority tasks need to be considered.…”

“…Significant research has been done on hierarchical scheduling [14,16,17,28,30]. In hierarchical scheduling, an interface specifies how processor resources are provided over time, e.g., "a server provides capacity of 5 time units at a period of 20 time units."…”