Thesis supervisor: Balázs Sonkoly
Location of studies (in Hungarian): Department of Telecommunications and Media Informatics Abbreviation of location of studies: TMIT
Description of the research topic:
"Serverless architectures and microservices provide a novel way of
software development and operation. A new generation of applications
with features never seen before is promised, while the burden on
developers and application providers is reduced or more exactly,
shifted towards the cloud operators. For example, web applications
typically follow the microservice architecture, where the monolithic
software is broken down into smaller independently managed components
usually realized by software containers. Allocating resources
dynamically to constituent containers and scaling them properly on
demand can be a challenging task, and on-demand vertical and
horizontal resource scaling in an arbitrary scale are essential
features to be provided inherently by future cloud platforms. The
scaling logic can be driven by various service management goals, e.g.,
either minimizing resource usage while sustaining a given service
quality target, or minimizing Service Level Agreement (SLA) violations
no matter the price on the side of provisioned resources. The cloud
scaling decision-making is further aggravated by the peculiar scaling
behavior of the managed application, i.e., the function that
translates the amount of requests to be served respecting the given
SLA constraints to the necessary amount of resources to provision.
Moreover, latency sensitive applications with strict delay constraints
pose additional challenges to the underlying platforms and novel
mechanisms are required.
The goal of the thesis is to provide enhanced and generally applicable
solutions for modeling, operating and optimizing microservice based
applications on serverless platforms.
The main goals are the following:
- Propose novel AI/ML models for microservice based applications
capturing compute and network resource requirements (resource
profiles) and end-to-end latency characteristics
- Propose/Adopt AI/ML techniques to learn serverless cloud platforms
characteristics
- Design and implement novel resource scaling algorithms, including
horizontal, vertical and combined ones, leveraging the novel
application and platform models
- Define novel techniques and implement corresponding controller
software to optimize the operation of microservices on different
serverless platforms
"
Required language skills: English Number of students who can be accepted: 1
Login
