KRATER Benchmark Dataset

Raw benchmark results from KRATER (Kubernetes Runtime Analysis Tool for Empirical Research), a benchmarking framework comparing WebAssembly and Linux container runtimes in K3s. Three datasets are included, corresponding to GCC/WASI-SDK optimization levels O0, O2 and O3. Each JSON file contains per-trial data for one runtime environment: throughput (MFLOPS), CPU utilization, memory usage, cold start latency and throttling events, collected across 30 trials at matrix sizes N = 256 and N= 1024. Hardware: AMD Ryzen 9 8940HX, Pop!_OS 24.04, K3s with v0.6.0 runwasi shims for Wasmtime, Wasmer and WasmEdge.