Current Theses

Title

Student

Supervisor(s)

Description
AFCL Environment – Development and Scalable Execution of Portable Function Choreographies Across Multiple Serverless Cloud Platforms Stefan Pedratscher Sashko Ristov details
VolatileSim: A simulation framework for cloud volatile resources
 Christoph Schöpf Sashko Ristov details

Title AFCL Environment – Development and Scalable Execution of Portable Function Choreographies Across Multiple Serverless Cloud Platforms
Language Englisch
Supervisors Sashko Ristov
Student Stefan Pedratscher
Description This master thesis designs an abstract function choreography language (AFCL) and JAVA API (AFCLCore) to build serverless workflow applications or function choreographies (FCs) at a high-level of abstraction. AFCLCore will be integrated in several tools that resulted by bachelor theses. Such AFCLCore-based tools will be integrated into the AFCL environment, which will be able to develop and run portable function choreographies (FCs) across multiple serverless cloud platforms by overcoming individual FaaS provider limits.
Tasks
  • Contribute to development of Dependency-Aware Faasifier (DAF) and evaluate it with the state-of-the-art FaaSifiers.
  •  Define a Java Interface jFaaS to facilitate portable serverless application development.
  •  Describe FCs at a high-level of abstraction in AFCL and create a Java application programming interface to easily build, parse, write and validate FCs.
  •  Integrate AFCLCore in the multi-FaaS enactment engine
  •  Define a model within the AFCL runtime system for distributing parallel loops of an FC across multiple FaaS cloud providers using jFaaS and AFCLCore.
  •  Integrate fault tolerance in AFCL.
  •  AFCL Environment: Integrate AFCLCore-based FCEditor, Scheuler and enactment engine.
Theoretical skills Distributed Systems, workflow applications, serverless computing
Practical skills Java, JavaScript, Web development, YAML, JSON, SDK for public cloud providers
Additional information enactment engine, Scheduler and FCEditor is provided.

Title VolatileSim: A simulation framework for cloud volatile resources
Language Englisch
Supervisors Sashko Ristov
Student Christoph Schöpf
Description Volatile cloud instances (e.g. Amazon’s Spot Instances) are cheaper than On-Demand instances with the same resources. However, they are prone to failures because the cloud provider can reclaim them at any time. Many parameters that can describe and model the volatile resources, such as frequency of no-capacity interruptions, fulfilment rate, waiting time to fulfil the request, cannot be predicted, which hardens the researchers to simulate such a dynamic environment with high accuracy. The goal of this master thesis is to analyse the behavior of various cloud resources, all events of their life-cycle. Based on the analysis, the student should develop a simulation framework VolatileSim that can simulate the realistic behavior of volatile cloud resources and running complex workflow applications.
Tasks
  • Becoming familiar with the volatile cloud resources
  • Develop a simulation framework VolatileSim
  • Evaluate the accuracy of VolatileSim against the real execution in cloud
  • Evaluate the how various volatile cloud resources parameters affect the makespan and overall cost of scientific workflow execution
Theoretical skills
  • Distributed Systems
  • Fault tolerance
  • Cloud IaaS
  • Scientific workflows
Practical skills
  • Good Java knowledge, EC2 SDK