I.J. Information Technology and Computer Science, 2017, 6, 59-66 Published Online June 2017 in MECS (http://www.mecs-press.org/) DOI: 10.5815/ijitcs.2017.06.08 Copyright © 2017 MECS I.J. Information Technology and Computer Science, 2017, 6, 59-66 Optimization of System’s Performance with Kernel Tracing by Cohort Intelligence Aniket B. Tate Dept. of Computer Engineering, Vishwakarma Institute of Information Technology, Pune, 411048, India E-mail: t.aniket.t@gmail.com Laxmi A. Bewoor Dept. of Computer Engineering, Vishwakarma Institute of Information Technology, Pune, 411048, India E-mail: laxmiabewoor@gmail.com Abstract—Linux tracing tools are used to record the events running in the background on the system. But these tools lack to analyze the log data. In the field of Artificial Intelligence Cohort Intelligence (CI) is recently proposed technique, which works on the principle of self- learning within a cohort. This paper presents an approach to optimize the performance of the system by tracing the system, then extract the information from trace data and pass it to cohort intelligence algorithm. The output of cohort intelligence algorithm shows, how the load of the system should be balanced to optimize the performance. Index Terms—Kernel Trace, Linux Tracing Tool Next Generation (LTTng), Metaheuristics, Cohort Intelligence. I. INTRODUCTION Kernel is the black box of the operating system which handles the system resources like Memory, Input Output, CPU time etc. Observing system and understanding what is happening inside the operating system is the important aspect to optimize the system performance. Kernel tracing facilitates to demonstrate various activities running inside the Operating System. Kernel tracing is the activity of listing down, which OS process is using which system resources at what time. There are various tools available to trace down kernel activities like LTT, LTTng, DTrace, FTrace etc. these tracing tools provides details about processes running in the background and their resource uses. This traced information can assist performance optimization of operating system. These tracing tools provide details about process and resources they use but lack to analyze log data and extract knowledge from it. [1] Another approach to improve the system’s performance is to optimize the scheduler. Scheduler is one of the major components which affect system performance largely. The aim of the scheduling is to share the resources by a number of processes. Scheduling is central to an Operating-system’s design and constitutes an important topic in the computer science curriculum. Heuristics like Genetic Algorithm (GA) [2][3], Ant Colony Optimization (ACO) [4][5] have been implemented on the scheduler to optimize system performance like maximize resource utilization, minimal execution time etc. which shows better results. While scheduling a process on one of the cores, the scheduler considers the average waiting time, turnaround time, time quantum for a process, number of context switches, earliness, the tardiness of process etc. But the scheduler does not take CPU load into consideration. As a result of this, the cores get unevenly loaded and many of the cores will be kept in ideal state. Cohort Intelligence (CI) is recently introduced meta-heuristics [6][7] it works on self-supervised learning behavior in a cohort. Cohort refers to a group of candidates which communicate with each other to achieve the common goal of the cohort by improving individual’s behavior. In the cohort, every candidate observes every other candidate and tries to adopt a new behavior. Each candidate must follow other candidate or itself in the cohort, which results in improvement of behaviors of the candidate. In this way, every candidate helps in improving the cohort’s behavior. The cohort is considered to be saturated if the behavior of candidate within adjacent iterations is nearly equal, a maximum number of iterations occur [6][7]. This paper proposes an approach to use the LTTng 2.7 tool for tracing out the kernel (Ubuntu12), analyze the trace data and pass CPU load information to CI algorithm. CI algorithm works on that data and gives a set of cores as an output to reschedule the processes. Finally, a simulation is shown how to migrate process on core or set of cores to balance system’s load. The remainder of this paper is organized as follows. Section II covers literature survey regarding kernel tracing, system optimization techniques, and cohort intelligence. In section III, steps to trace Linux system (Ubuntu12) are listed. Section IV gives an overview of system architecture. Section V gives flow chart of CI algorithm and explanation of how CI is used in the proposed system. Section VI shows a mathematical model of proposed system. The result and analysis of the proposed system are presented in section VII. The concluding remark is presented in section VIII of the paper. II. RELATED WORK Anderson et al. [1] suggested software performance