High-Density Dynamic Website Hosting: A Reproducible Evaluation of the Kooboo Architecture

This technical report evaluates the density and end-to-end request handling of a multi-site web architecture. A single Kooboo process hosted 5,000 separately created and independently editable dynamic websites on a 2-vCPU, 4-GB public cloud server. A separate cloud server sent 90,000 HTTPS requests across all 5,000 hostnames over the public Internet at a configured rate of 150 request starts per second. Each complete response body was downloaded and checked for a site-specific marker. The run produced 89,969 verified responses, 31 connection-establishment timeouts, no HTTP errors, and no content mismatches. The measured first-attempt success rate was 99.97%, with 146.59 request starts per second. Complete-response latency was 296.73 ms at p50, 1,431.27 ms at p95, and 3,393.82 ms at p99. The Kooboo process consumed an average of 45.6% of the total two-vCPU capacity and reached an approximately 2.50-GiB sampled peak resident set size. The websites used one controlled dynamic blog package. Each site contained 10 blog records and an uncached home-page content query. No full-page output cache was used. This result demonstrates the measured capacity of this controlled workload. It should not be interpreted as evidence that every possible collection of 5,000 production websites would behave identically. The methodology, request-level data, server monitoring records, C# test-client source, binaries, checksums, site package, and reproduction instructions are publicly available from the related GitHub repository. This is a first-party technical report and has not been peer reviewed.