A Workflow for the Integral Performance Analysis of Cloud Applications Using Monitoring and Tracing Techniques

Publikation: Beitrag in Buch/Konferenzbericht/Sammelband/GutachtenBeitrag in KonferenzbandBeigetragenBegutachtung


Considering the cost effectiveness, elasticity, and flexibility of virtualized cloud environments, porting HPC applications to those environments and executing them within these settings is becoming more and more popular. For this purpose, traditional HPC applications have to be redesigned as cloud applications. An analysis of the performance of the redesigned applications within the cloud environment is in-dispensable, if the applications should be efficiently executed in the cloud environment. This paper proposes a workflow for the integral performance analysis of cloud applications within a cloud environment using monitoring and tracing techniques. For this, collectd acts as a lightweight monitoring daemon for recording performance data from outside of the applications, Score-P as a profiling and tracing tool for recording the performance data from the inside. Thus, this workflow will help in answering the question "How and why an application behaves like this within the cloud environment?". In order to show the usability of the proposed workflow, a parallel client server application was selected and adapted for the execution in a private OpenStack cloud. Performance measurements of the example running in the cloud environment could be successfully done according to the proposed workflow. In particular, performance metrics from both the outside and the inside of the application could be obtained to analyze and evaluate the performance of the application in detail.


TitelICCBDC 2017: Proceedings of the 2017 International Conference on Cloud and Big Data Computing
ErscheinungsortNew York, NY, USA
Herausgeber (Verlag)Association for Computing Machinery (ACM), New York
ISBN (Print)978-1-4503-5343-4
PublikationsstatusVeröffentlicht - 1 Sept. 2017


ReiheICCBDC: International Conference on Cloud and Big Data Computing

Externe IDs

Scopus 85045726410
ORCID /0000-0003-3649-2433/work/141544952


Forschungsprofillinien der TU Dresden


  • Cloud application, collectd, Container, Docker, Micro services, monitoring, performance analysis, Score-P, tracing, workflow