Research
All four projects measure infrastructure on test setups I built, alone or with colleagues. The newest injects faults into an 11-VM Kubernetes cluster to see which guarantees hold. With Vedran Dakić, Jasmin Redžepagić and Luka Žgrablić we measured what four CNI plugins add to throughput and latency, and with Jurica Slovinac we ran the same benchmarks on the bare host and in Kubernetes pods. My graduate thesis compared what ten cloud services charge with how fast they respond.
Pages on published work link to the papers; the Kubernetes programme is still in preparation, with no results yet.
How Kubernetes behaves when its components disagree
Fault-injection experiments on an 11-VM Kubernetes lab, asking which guarantees still hold when the control plane and the nodes report different things about the same cluster. Three papers are in preparation.
Network performance of Kubernetes CNI plugins
We built Ansible automation that installs a Kubernetes CNI plugin, tunes the hosts and measures bandwidth and latency against the same network without Kubernetes, then used it to compare Antrea, Calico, Cilium and Flannel. Two 2024 papers in MDPI journals report the method and the results on 10 Gbit/s links. A 2026 paper in Electronics takes the comparison to 100 Gbit/s Ethernet and InfiniBand, adds pod-level RoCE, and shows that on multi-homed nodes the interface carrying the node addresses decides which link CNI traffic uses.
Kubernetes overhead on bare metal
The same CPU, memory and disk benchmarks, run on three x86 servers and two Raspberry Pis both directly on the host and as Kubernetes CronJobs, plus a prototype scheduler that places pods by past benchmark results. The data fed a 2024 paper on workload placement in Kubernetes-based HPC data centres.
Comparing the price-to-performance ratio of serverless services in the public cloud
My 2023 graduate thesis. I ran the same test website and web API on ten cloud services to find which serverless platform gives the best performance for the price.