Utility of Univariate Forecasting for Workload Metrics Predictions in Enterprise Applications

Andrey Kharitonov, Roheet Rajendran, Hendrik Müller, Klaus Turowski

2023

Abstract

Modern enterprise IT systems are complex solutions that require careful planning of computational capacities and placement, especially in the cloud environments where the total cost of ownership directly depends on provisioned resources. The decision process on infrastructure transformation or capacity sizing of existing IT landscapes can be supported by collecting and analyzing the workload data of the running systems. However, the scope and length of this data are limited, as its collection is often an expensive and lengthy process. Therefore, within this work, we empirically evaluate multiple techniques for extending the workload data by employing various univariate time series forecasting algorithms. We analyze a use case of SAP-based enterprise applications and rely on real-world workload data collected from various running SAP system landscapes. Our analysis demonstrates that XGBoost is best suited for univariate forecasting SAP-specific key performance indicators for both stationary and trending time series. However, the shape of the workload profile has a high degree of influence on the results of the forecasting. Enterprise applications’ workload data that represent regular day-to-day operations without irregular events is a prerequisite for accurate forecasting.

Paper Citation

in Harvard Style

Kharitonov A., Rajendran R., Müller H. and Turowski K. (2023). Utility of Univariate Forecasting for Workload Metrics Predictions in Enterprise Applications. In Proceedings of the 15th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 3: KMIS; ISBN 978-989-758-671-2, SciTePress, pages 231-240. DOI: 10.5220/0012206600003598

in Bibtex Style

@conference{kmis23,
author={Andrey Kharitonov and Roheet Rajendran and Hendrik Müller and Klaus Turowski},
title={Utility of Univariate Forecasting for Workload Metrics Predictions in Enterprise Applications},
booktitle={Proceedings of the 15th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 3: KMIS},
year={2023},
pages={231-240},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012206600003598},
isbn={978-989-758-671-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 15th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management - Volume 3: KMIS
TI - Utility of Univariate Forecasting for Workload Metrics Predictions in Enterprise Applications
SN - 978-989-758-671-2
AU - Kharitonov A.
AU - Rajendran R.
AU - Müller H.
AU - Turowski K.
PY - 2023
SP - 231
EP - 240
DO - 10.5220/0012206600003598
PB - SciTePress