Back to overview

Evaluating Multi-tenant Live Migrations Effects on Performance

Type of publication Peer-reviewed
Publikationsform Proceedings (peer-reviewed)
Author Rosinosky Guillaume, Labba Chahrazed, Ferme Vincenzo, Youcef Samir, Charoy Françcois, PautassoCesare,
Project Declarative Continuous Performance Testing for Microservices in DevOps
Show all

Proceedings (peer-reviewed)

Page(s) 61 - 77
ISBN 978-3-030-02610-3
Title of proceedings On the Move to Meaningful Internet Systems. OTM 2018 Conferences

Open Access


Multitenancy is an important feature for all Everything as a Service providers like Business Process Management as a Service. It allows to reduce the cost of the infrastructure since multiple tenants share the same service instances. However, tenants have dynamic workloads. The resource they share may not be sufficient at some point in time. It may require Cloud resource (re-)configurations to ensure a given Quality of Service. Tenants should be migrated without stopping the service from a configuration to another to meet their needs while minimizing operational costs on the provider side. Live migrations reveal many challenges: service interruption must be minimized and the impact on co-tenants should be minimal. In this paper, we investigate live tenants migrations duration and its effects on the migrated tenants as well as the co-located ones. To do so, we propose a generic approach to measure these effects for multi-tenant Software as a Service. Further, we propose a testing framework to simulate workloads, and observe the impact of live migrations on Business Process Management Systems. The experimental results highlight the efficiency of our approach and show that migration time depends on the size of data that have to be transferred and that the effects on co-located tenants should not be neglected.