k8s-cluster

Kubernetes Cluster @ UFRGS

This cluster is a heterogeneous, research-oriented infrastructure designed to support advanced experiments in distributed computing, networking, and cloud-native applications. Managed by the Institute of Informatics at UFRGS, it integrates diverse computing resources ranging from high-performance servers to edge devices.

Infrastructure Highlights:

Computing Power: A mix of Intel Xeon servers, desktop nodes, and ARM64 Edge devices (Raspberry Pi 4), totaling over 600 CPU cores and 1.4 TB of RAM.
Storage: 59 TB of total storage (SSD & HDD) managed via Longhorn for persistent and block storage.
Networking: Software-defined networking powered by Calico CNI and OpenFlow-native switches (Centec v350/v330), with future expansion plans for 5G USRP setups.

Software Stack:

The cluster employs a robust cloud-native stack to facilitate orchestration, observability, and service management:

Orchestration & Service Mesh: Kubernetes with Istio (Ambient mode), Envoy, and Kiali.
Serverless: Knative for event-driven applications and auto-scaling.
Observability: A comprehensive monitoring suite including Prometheus, Grafana, Loki, and Jaeger.

Research Focus:

The infrastructure currently supports over 25 active members and key projects in 5G programmability (PORVIR-5G), Digital Twins (PETWIN), and Industrial AI (OntoKG, Nautilus Twin).

For a complete technical deep-dive into the cluster’s architecture, software configurations, and expansion plans, please refer to the following Google Drive presentation:

View detailed cluster software and systems information here

Computation Resources (current)

We now have a public app running on the cluster that shows some relevant information on the hosts currently connected to it:

View detailed node information here

Cluster Public Dashboards

We provide a few public Grafana dashboards with an overview of the main metrics monitored with Prometheus.