Robust fuzzy CPU utilization control for dynamic workloads

Abstract

a b s t r a c tIn a number of real-time applications such as target tracking, precise workloads are unknown a priori but may dynamically vary, for example, based on the changing number of targets to track. It is important to manage the CPU utilization, via feedback control, to avoid severe overload or underutilization even in the presence of dynamic workloads. However, it is challenge to model a real-time system for feedback control, as computer systems cannot be modeled via physics laws. In this paper, we present a novel closedloop approach for utilization control based on formal fuzzy logic control theory, which is very effective to support the desired performance in a nonlinear dynamic system without requiring a system model. We mathematically prove the stability of the fuzzy closed-loop system. Further, in a real-time kernel, we implement and evaluate our fuzzy logic utilization controller as well as two existing utilization controllers based on the linear and model predictive control theory for an extensive set of workloads. Our approach supports the specified average utilization set-point, while showing the best transient performance in terms of utilization control among the tested approaches.

Computer State Evaluation Using Adaptive Neuro-Fuzzy Inference Systems

“…We assumed that the utilization of cache, memory, storage, and bus includes all the main features affecting the infrastructure of the computing system, namely the CPU utilization. We used adaptive neuro-fuzzy logic to determine the evaluation rule strength by enhancing our previous work [ 6 ]. This ANFIS is useful for any user who has an elementary knowledge and is just studying performance and utility of computer components.…”

“…However, these assessment techniques are based on some datasets that cause some inaccuracy during the implementation in the particular systems and on analyzing existing real systems. Additionally, there are fuzzy models for predicting the utilization of CPU based on trends and previous states of the workload [ 6 , 7 ] or taking into account historical data of the central processor time, RAM (read time, write time, swap time), the throughput of I/O and bus. In addition, these objects have many pointers (utilization, time, throughput, latency, delay, speed, frequency, and percentage), which confuses the assessing and prediction of computer components.…”

“…For the purpose of predicting CPU utilization, Bey et al [ 21 ] combined ANFIS with the clustering procedure employed on CPU utilization time series. The more pertinent research is found in [ 6 ], where the authors applied fuzzy logic control theory to modify CPU workloads as needed such that use converges with a specified setting even for dynamic workloads. The study stated that fuzzy logic control is unrelated to a mathematical model of a controlled system or operating range, in contrast to proportional, integral, and differential (PID) approaches and other predictive models.…”

“…The probability of introducing design flaws as a result of statistical inaccuracies included in the black box installation model is small because there is no fuzzy logic controller model. By conducting a logical study of the nonlinear relationship between utilization and load variations, Basaran and his research colleagues suggested a new fuzzy closed-loop system for managing CPU use [ 6 ].…”

Adaptive load shedding via fuzzy control in data stream management systems

“…Basaran et al [25] show that fuzzy control considerably outperforms a PI controller and a model predictive controller in terms of CPU utilization control in a real-time operating system. However, these approaches do not aim to support load shedding considering distributed DSMS semantics discussed in Section I.…”