Artikel in Tagungsband INPROC-2019-36

Yussupov, Vladimir; Breitenbücher, Uwe; Hahn, Michael; Leymann, Frank: Serverless Parachutes: Preparing Chosen Functionalities for Exceptional Workloads.
In: Proceedings of the 2019 IEEE 23rd International Enterprise Distributed Object Computing Conference (EDOC 2019).
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik.
S. 226-235, englisch.
IEEE Computer Society, Oktober 2019.
DOI: 10.1109/EDOC.2019.00035.
Artikel in Tagungsband (Konferenz-Beitrag).
CR-Klassif.D.2.2 (Software Engineering Design Tools and Techniques)
D.2.11 (Software Engineering Software Architectures)
D.3.4 (Programming Languages Processors)
KeywordsServerless; FaaS; Function-as-a-Service; Scalability; Failover; Annotation

Function-as-a-Service (FaaS) is an emerging cloud service model that enables composing applications using arbitrary, small, and event-driven code snippets managed by cloud providers and that can be scaled to zero. The scalability properties of FaaS look attractive for handling rare or unexpected high loads that affect only particular functionalities of the application. However, deciding on the component granularity upfront or reengineering the architecture of an entire application for rare workloads is often a very difficult challenge or even infeasible. In this work, we introduce a method that prepares annotated functionalities for handling rare workloads by automatically extracting them from the source code of the application and additionally deploying them as FaaS functions, while keeping the original application's functionalities and architecture unchanged. In this way, the benefits of FaaS can be leveraged without the need to reengineer the application only for rare cases. We validate our method by means of a prototype, evaluate its feasibility in a set of experiments, and discuss limitations and future work.

Volltext und
andere Links
EDOC 2019
KontaktVladimir Yussupov
Abteilung(en)Universität Stuttgart, Institut für Architektur von Anwendungssystemen
Eingabedatum16. Januar 2020
   Publ. Institut   Publ. Informatik