Master's Thesis

Performance Isolation in GraalVM Native Image Isolates

Filipe Gomes Crispim de Sousa — 2022

Key information

Authors:

Filipe Gomes Crispim de Sousa (Filipe Gomes Crispim de Sousa)

Supervisors:

Rodrigo Fraga Barcelos Paulus Bruno (Rodrigo Fraga Barcelos Paulus Bruno); Luís Manuel Antunes Veiga (Luís Manuel Antunes Veiga)

Published in

November 23, 2022

Abstract

In the cloud computing service model Function-as-a-Service (FaaS), small, stateless, and event-driven functions, are invoked countless times in parallel. For each of these function invocations, a new container and runtime will have to be started, resulting in non-negligible latency. One solution to this problem is by co-locating functions in the same runtime, which reduces the amount of runtime start ups and improves the memory footprint, since there are less runtimes. Since function invocations will share the runtime, there is no control over how much CPU a function gets in relation to others. If a function has multiple threads it may grab a bigger slice of the CPU. In this work, we design and implement a mechanism for managing the CPU shares between co-located functions in a FaaS environment, in the form of an http server that receives requests to run client's functions. The main purpose is to allow cloud computing clients to set CPU requirements for their functions and making sure these are met to the best of the system's abilities.

Publication details

Authors in the community:

Supervisors of this institution:

Fields of Science and Technology (FOS)

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

Publication language (ISO code)

eng - English

Rights type:

Embargo lifted

Date available:

September 13, 2023

Institution name

Instituto Superior Técnico