Thesis on Load Balancer Algorithm for Job scheduling in Software

At the Propulsion Software Factory we run a Continuous Integration & Delivery environment for about 500 software developers within the Volvo group. The Software developed and deployed through the Factory represents the technical areas of Electromobility and Conventional Powertrain.
Today in our system we are running a large variety of jobs on a number of build servers. Our Jenkins machine is allocating these jobs on different builders with a simple strategy to prefer execution on the same machine as ran the job last time. This leads to non-optimal usage of our build capacity. This reduced throughput of the Factory.


Suitable background


Computer Science, IT, Mathematics or Physics

Description of thesis work


The thesis should investigate the possibility to introduce a smart algorithm that can analyze the results, run time and logs of all jobs run on all builders and allocate jobs to builders optimizing the complete job queue and limit builder-related errors. The algorithm should be self-learning and adaptive to infrastructure changes and builder performance over time.


Expected results:
An algorithm introduced on our Jenkins Master for job scheduling that evolves overtime.
Comparison of job throughput with and without the algorithm.
A discussion of the advantages with smarter job scheduling and continued improvements

Thesis Level: Master

Language: English or Swedish


Starting date: 2023-02-01


Number of students: 1-2


Tutor
Jim Fischer, Group Manager Software Factory, +46 313231697
jim.fischer@volvo.com