MÓDULO K6.Art¶

K6 agents — wire-level TCP/TLS load with p50/p95/p99 stats.

Function¶

grafana/k6 binary running HTTP/1.1, HTTP/2, HTTP/3 scenarios. Measures wire-level metrics: requests-per-second, error rate, p50/ p95/p99 latency, data_received bytes.

Differentiator vs PW.Art: K6 measures datasheet-style throughput (no browser overhead). PW measures real-world Web Vitals. Side-by- side reveals the "inflation gap" (datasheet vs production).

Identity¶

Element	Value
Plane	DATA
Internal code	`k6-agent`
K8s namespace	`web-agents`
OOBI slot	n/a (data-plane only)
Network	VLAN 30 (K6 TRUST), subnet `172.17.0.0/16`

Dependencies¶

grafana/k6 base image
Multus + macvlan for VLAN 30

Operator controls¶

/admin/k6/agents — register / status
VU (virtual user) count, ramp profile, duration
HTTP version per scenario (h1 / h2 / h3)

Key telemetry¶

k6_http_req_duration_p50_ms, _p95_ms, _p99_ms
k6_http_reqs_per_sec
k6_http_req_failed_rate
k6_data_received_bytes
k6_iterations — VU iterations completed

Notes¶

K6 saturates much higher RPS than PW per-host (no browser overhead). Typical scale: 1 K6 agent = ~10K RPS; 1 PW agent = ~100 RPS.

Memory: project_pw_vs_k6_strategic_differentiation_2026_05_10
PW.Art — companion measurement engine
Open Q: K6 vs iperf3 baseline (per discuss_vpn_ipsec_simulation, resolved by Cloud Endpoint Service ADR 0023 — iperf3 standardized)