Research & Case Studies

Selected technical findings from our continuous distributed telemetry testbeds.

STUDY // 2026-A

Congestion Window Dynamics under Artificial Jitter

Comparing BBRv3 and Cubic behavior under sustained 4% packet drop rates across high-latency Eurasian fiber routes. Analysis of queue recovery intervals and throughput collapse thresholds.

STUDY // 2026-B

Static Asset Delivery Efficiency: Self-Hosting vs CDN

Benchmarking cold-start TLS handshake overhead, HTTP/2 multiplexing limits, and real-world Time To First Byte (TTFB) comparing monolithic edge servers to multi-tenant CDN tiers.

STUDY // 2026-C

Bufferbloat Mitigation in Asymmetric Uplinks

Evaluation of CoDel and FQ-PIE queueing disciplines implemented at the host socket layer. Realized a 78% reduction in latency variance during saturated outbound transmissions.

Methodology & Verification Environment

All experiments are conducted across standardized hardware instances running Linux 6.x kernels with locked CPU frequency scaling and direct bare-metal network interface bindings.

Telemetry collection agents execute in isolated cgroups with strictly capped memory envelopes to prevent metric collection from perturbing active network benchmarks.