Research
Applied Computing Research Labs is dedicated to advancing the state of the art in distributed systems through rigorous, open-source research applied in practical settings. ACRL’s team, led by experts with strong academic credentials, has pioneered innovative approaches to solving complex computational challenges.
Here are some of our most exciting research projects:
Kompile
Kompile allows you to sidestep the “monolith vs. microservices” question; write a single program in Go and automatically “compile” it to run on Kubernetes. Goroutines are converted to microservice calls, and channel communications are converted to network requests. Joint work with Tim Goodwin at UCSC.
Cluster Autoscaler/Karpenter Comparison
An in-depth analysis comparing the performance of the Kubernetes Cluster Autoscaler and Karpenter, two competing node autoscaling solutions for Kubernetes.
Kube-scheduler-rs
A “reference” implementation for building a Kubernetes scheduler in Rust, using the kube-rs Kubernetes API bindings and controller utils.
prom2parquet and DataKube
Tools for collecting, storing, and analyzing Prometheus metrics from Kubernetes. prom2parquet is a remote-write
backend for Prometheus that saves time-series data to parquet files. DataKube is a DuckDB-backed collection of utility
functions for analyzing the collected data.
🔥Config
FireConfig is a prototype of a Python-based configuration-as-code library for generating Kubernetes manifests built on top of cdk8s.