Approximate Analysis of Real-Time Database Systems

During the past few years, several studies have been made on the performance of real-time database systems with respect to the number of transactions that miss their deadlines. These studies have used either simulation models or database testbeds as their performance evaluation tools. We present here a preliminary analytical performance study of real-time transaction processing. Using a series of approximations, we derive simple closed-form solutions to reduced realtime database models. Although quantitatively approximate, the solutions accurately capture system sensitivity to workload parameters and indicate conditions under which performance bounds are achieved. 1 I n t r o d u c t i o n In a broad sense, a real-time database system (RTDBS) is a transaction processing system that is designed to handle workloads where transactions have service deadlines. The objective of the system is to meet these deadlines, that is, to process transactions before their deadlines expire. Therefore, in contrast to a conventional DBMS where the goal usually is to minimize transaction response times, the emphasis here is on satisfying the timing constraints of transactions. Transactions may miss their deadlines in a realt ime database system due to contention for physical resources (CPUs, disks, memory) and logical resources (data). During the last few years, several detailed studies [1, 3, 5] have evaluated the performance of various real-time transaction resource scheduling policies with respect to the number of missed transaction deadlines. These studies have either used simulation models [1, 3] or used database testbeds [5] as their performance evaluation tools. The lack of analytical *This work was initiated while the author was with the Systems Research Center, Univ. of Maryland (College Park) and was supported in part by a SRC Post-Doctoral Fellowship. Input Transaction., R a t e )~ Killed ~t)~ l Transactions _LI. I Successful Transactions