← All capstone topics
Industry + research Edge and cloud scheduling

Budget- and deadline-aware autoscaling of Kubernetes microservices with reinforcement learning

Learn when to add or remove copies of each microservice so that response times stay under target while the cloud bill stays within budget.

Level: Undergraduate Format: Individual project Duration: 2 semesters Supervisor: Dr. Hamza Djigal

The problem

Web applications are split into many microservices running on Kubernetes. The default autoscaler reacts to CPU usage with fixed thresholds: it scales too late during bursts and wastes money at night. The right number of copies depends on the traffic pattern and on how the services call each other.

Who needs it

Every company running on Kubernetes (e-commerce, fintech, SaaS); cloud cost and latency targets (SLOs) are tracked by management.

Idea of the approach

  • Deploy a benchmark microservice application (DeathStarBench social network or the Google Online Boutique demo) on a small Kubernetes cluster (kind/minikube or cloud credits).
  • Replay realistic traffic traces.
  • Train a reinforcement-learning agent (DQN or PPO) whose reward combines latency violations and cost; compare with the Kubernetes HPA and a rule-based baseline.
  • Add a budget constraint as in BUDA/BUCAS.

What you will build and deliver

  • The autoscaler as a Kubernetes controller.
  • Experiments with latency and cost curves.
  • Paper and demo.

Skills needed

Docker, Kubernetes basics, Python, reinforcement learning (can be learned during the project).

Research contribution & where to publish

Contribution: budget-constrained RL autoscaling with explicit cost accounting. Venues: IEEE CLOUD, ICSOC, IEEE Transactions on Cloud Computing, Future Generation Computer Systems.

Key references — start reading here

  1. K. Rzadca, P. Findeisen, J. Swiderski, P. Zych, P. Broniek, J. Kusmierek, P. Nowak, B. Strack, P. Witusowski, S. Hand, J. Wilkes. Autopilot: Workload Autoscaling at Google. ACM EuroSys, 2020.
  2. H. Qiu, S. S. Banerjee, S. Jha, Z. T. Kalbarczyk, R. K. Iyer. FIRM: An Intelligent Fine-grained Resource Management Framework for SLO-Oriented Microservices. USENIX OSDI, 2020.
  3. Y. Gan, Y. Zhang, D. Cheng, A. Shetty, P. Rathi, N. Katarki, et al.. An Open-Source Benchmark Suite for Microservices and Their Hardware-Software Implications for Cloud & Edge Systems. ACM ASPLOS, 2019.
  4. H. Djigal, L. Liu, J. Luo, J. Xu. BUDA: Budget and Deadline Aware Scheduling Algorithm for Task Graphs in Heterogeneous Systems. IEEE/ACM IWQoS, 2022.

Interested?

Send the topic title and the courses you have taken.

Contact Dr. Hamza Djigal

Other topics in Edge and cloud scheduling