Construction of Proactive Monitoring Model using Forecasting Techniques in the SCOM Software Complex

Abstract

The majority of companies depend on their information systems, the stability of infrastructure operations and the failover of computing resources. Various monitoring tools are mostly used to automate the benchmarking process of company.

The company that has a large distributed infrastructure should pay close attention to this process, as it makes the state of operations difficult to maintain, and increases the probability of the loss of functionality for errors or even shutdown of some servers. The one of solutions is reactive monitoring.

Reactive monitoring is a technique where system administrators use monitoring tools to continuously collect data that determine the active and current status of information system environment. Measurements obtained from real-time monitoring tools illustrate the performance data of current information environment.

However, if we discuss the main metrics of system resources, such as the level of processor load, RAM or disk usage, their change can be quite fast. And for servers that are responsible for critical functions, the problem of full resource usage is important. This problem can be solved with proactive monitoring.

Proactive monitoring is a set of monitoring tools that not only collect real-time information, but also predict possible failures before they impact end users[1].

The purpose of this article is to choose methods of time series forecasting for the resources load that are going to be combined into a single hybrid method. The final solution will be used in the management pack of software complex System Center Operations Manager (SCOM) that is widely used by companies with large infrastructure[2]. The forecasting methods such as Least squares, SMA and EMA were considered in this work.