Using control theory to achieve service level objectives in performance management

Abstract

A widely used approach to achieving service level objectives for a software system (e.g., an email server) is to add a controller that manipulates the target system's tuning parameters. We describe a methodology for designing such controllers for software systems that builds on classical control theory. The classical approach proceeds in two steps: system identi®cation and controller design. In system identi®cation, we construct mathematical models of the target system. Traditionally, this has been based on a ®rst-principles approach, using detailed knowledge of the target system. Such models can be complex and dif®cult to build, validate, use, and maintain. In our methodology, a statistical (ARMA) model is ®t to historical measurements of the target being controlled. These models are easier to obtain and use and allow us to apply control-theoretic design techniques to a larger class of systems. When applied to a Lotus Notes groupware server, we obtain model-®ts with R 2 no lower than 75% and as high as 98%.In controller design, an analysis of the models leads to a controller that will achieve the service level objectives. We report on an analysis of a closed-loop system using an integral control law with Lotus Notes as the target. The objective is to maintain a reference queue length. Using root-locus analysis from control theory, we are able to predict the occurrence (or absence) of controller-induced oscillations in the system's response. Such oscillations are undesirable since they increase variability, thereby resulting in a failure to meet the service level objective. We implement this controller for a real Lotus Notes system, and observe a remarkable correspondence between the behavior of the real system and the predictions of the analysis. This indicates that the control theoretic analysis is suf®cient to select controller parameters that meet the desired goals, and the need for simulations is reduced.

Specification and management of QoS in imprecise real-time databases

“…We employ feedback control scheduling [10,11] to manage the quality of the service provided by the RTDB. The goal is to control the performance, defined by a set of controlled variables, such that the controlled variables satisfy a given QoS specification.…”

“…However they do not address the problem of maximizing QoS fairness among admitted tasks. In the work by Parekh et al, the length of a queue of remote procedure calls (RPCs) arriving at a server is controlled [11] using automatic control.…”

“…In earlier work [2,3] we presented the algorithms FCS-IC-1, FCS-IC-2, FCS-HEF, and FCS-HEDF for managing QoS using imprecise computation [9] and feedback control scheduling [10,11]. An important feature of this class of algorithms is their ability to adapt to various workloads and tolerate inaccurate estimates of execution times, and still conform to a given QoS specification.…”

Optimizing Quality of Service Using Fuzzy Control

“…Typically, the feedback controller is used for regulatory control, that is, to keep the performance metric at a desired reference value. Some examples are controlling queue length in Lotus Notes [1], buffer length in Internet routers [2], CPU and memory utilizations in Apache [3], and response times for web service differentiation [4]. Our interest is in generic approaches to optimizing the setting of tuning parameters so as to improve quality of service (e.g., minimize response times).…”

Introduction to Control Theory And Its Application to Computing Systems

“…More details on applying control theory to computing systems can be found in [16]. We describe the essentials of control design using the IBM Lotus Domino Server in [26]. The feedback loop is depicted in Figure 7.3.…”