2024 : 2 : 26
maryam Amiri

maryam Amiri

Academic rank: Assistant Professor
ORCID: https://orcid.org/0000-0002-7411-9552
Education: PhD.
ScopusId: 57146848900
Faculty: Engineering
Address:
Phone: 32625522

Research

Title
An online learning model based on episode mining for workload prediction in cloud
Type
JournalPaper
Keywords
Cloud computing Prediction Application Workload Episode mining Online learning
Year
2018
Journal Future Generation Computer Systems
DOI
Researchers Raffaela Mirandola ، Leyli Mohammad-Khanli ، maryam Amiri

Abstract

The resource provisioning is one of the challenging problems in the cloud environment. The resources should be allocated dynamically according to the demand changes of the applications. Over-provisioning increases energy wasting and costs. On the other hand, under-provisioning causes Service Level Agreements (SLA) violation and Quality of Service (QoS) dropping. Therefore the allocated resources should be close to the current demand of applications as much as possible. Thus, the prediction of the future workload of applications is an essential step before the resource provisioning. In our previous work, we proposed a Prediction mOdel based on SequentIal paTtern mINinG (POSITING), which considers the correlation between different resources and extracts behavioural patterns of applications independently of the fixed pattern length explicitly. Although POSITING provides reliable results, it is not able to adapt according to the workload variations. The application behaviour might change and drift due to the dynamic nature of cloud. For this purpose, we investigate the capabilities of online learning for POSITING. This paper proposes a Prediction mOdel based on epIsode miNing with the capabiliTy of onlIne learNinG (RELENTING) based on POSITING. Thus, in addition to the accuracy, adaptability, one of the most important characteristics of the application prediction models, is fulfilled. The performance of the proposed model is evaluated based on both real and synthetic workloads. The experimental results show that the proposed model adapts to the behavioural changes of the application and learns the new behavioural patterns rapidly in comparison to the other state-of-the-art methods such as moving average, linear regression, neural networks and hybrid prediction approaches.