Stateless node rebalancing and stateful building blocks for limitless autoscaling
Varadharajan, Thrivikraman
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/129548
Description
Title
Stateless node rebalancing and stateful building blocks for limitless autoscaling
Author(s)
Varadharajan, Thrivikraman
Issue Date
2025-04-20
Director of Research (if dissertation) or Advisor (if thesis)
Gupta, Indranil
Department of Study
Siebel School Comp & Data Sci
Discipline
Computer Science
Degree Granting Institution
University of Illinois Urbana-Champaign
Degree Name
M.S.
Degree Level
Thesis
Keyword(s)
autoscaling microservices
cpu limits
node rebalancing
stateful autoscaling
Language
eng
Abstract
Management of compute resources for cloud-native microservices relies heavily on autoscalers. Most autoscalers are built atop the fundamental mechanism of adjusting CPU limits---restricting the amount of CPU resources a service is allowed to use, and then they innovate scaling policies within those constraints. However, we show that the presence of CPU limits causes resource wastage and complicates autoscaler design. Thus, we advocate for the removal of CPU limits for allocating resources. Our design of Yet Another AutoScaler (YAAS) shows how this "limitless" design pathway opens up new challenges and opportunities. The development of YAAS was a collaborative effort, and this thesis focuses on addressing one specific challenge - the increase of CPU utilization beyond the allocation resulting in issues like node congestion and imbalances. To counter this, YAAS intelligently combines node scaling and rebalancing policies to meet SLOs (latency thresholds) with minimal disruptions. Experiments show that YAAS reduces CPU allocations by an average of 28% against limit-less baselines while satisfying SLOs. YAAS is limited to autoscaling stateless services. Although autoscalers for stateful services exist, most are paid solutions that are tailored to specific stateful services and do not provide users with full control. In this context, we present an in-depth study on scaling Redis and Kafka and identify the common building blocks for scaling stateful services.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.
