Skip to main content

Akamai CDN and Server-Timing Integration

ServiceRadar parses W3C Server-Timing metrics and Akamai EdgeScape headers into first-class typed columns on RumBeaconV1 and scalar metrics on metrics.rum.>. This populates the Edge vs Origin RTT vs Browser Render and Akamai CDN Cache Status & Edge Latency panels on the Web RUM dashboard without regex or JSON extraction.

1. Front your website with Akamai Delivery (Property Manager)​

In Akamai Control Center for each website property (for example shop.example.com):

  1. Origin Server: Point the Akamai property to your Kubernetes Envoy Gateway or origin hostname.

  2. Emit W3C Server-Timing headers: Configure Property Manager (via Modify Outgoing Response Header or EdgeWorkers) to emit Akamai's Server-Timing header on document responses:

    Server-Timing: cdn-cache; desc=HIT, edge; dur=12, origin; dur=45, ak_p; desc=iad-01

    ServiceRadar's Rust RUM collector (rust/otel/src/rum/scrub.rs) extracts these entries automatically:

    Server-Timing EntryStored Beacon ColumnScalar Metric (metrics.rum.>)
    cdn-cache; desc=HIT|MISS|REVALIDATEDcdn_cache_status--
    edge; dur=<ms>cdn_edge_rtt_msrum.akamai_edge_rtt_ms
    origin; dur=<ms>cdn_origin_rtt_msrum.akamai_origin_rtt_ms
    ak_p; desc=<pop> or cdn-pop; desc=<pop>cdn_pop (sets cdn_provider = "akamai")--

    If cross-origin subresources also emit Server-Timing that you want visible to Boomerang, add:

    Timing-Allow-Origin: https://shop.example.com
  3. Caching: Cache fingerprinted static assets (/assets/*, /images/*) aggressively at the edge and pass through or short-cache HTML pages so you observe both HIT and MISS origin RTT splits.

  4. DNS & TLS: Point the site's DNS record to your Akamai Edge Hostname (*.edgekey.net / *.edgesuite.net) with TLS provisioned via Akamai CPS.

2. Optional: Front /rum/v1/beacon with Akamai for EdgeScape Geo/ASN​

Even if /rum/v1/beacon is served directly from your cluster's Envoy Gateway, you receive full Core Web Vitals and website Server-Timing metrics (Edge RTT, Origin RTT, Cache HIT/MISS, POP).

If you also front your beacon hostname (rum.example.com) with Akamai:

  1. Enable True Client IP (True-Client-IP / X-Forwarded-For) and EdgeScape (Akamai-Edgescape carrying country_code, region_code, city, asnum, network) in Property Manager.

  2. Configure logCollector.otlp.rum.trustedProxies and logCollector.otlp.rum.cdnPeers in Helm values so the collector trusts Akamai-Edgescape only from your gateway and Akamai edge CIDRs:

    logCollector:
    otlp:
    rum:
    enabled: true
    trustedProxies:
    - "10.0.0.0/8"
    - "172.16.0.0/12"
    cdnPeers:
    - "198.51.100.0/24"

    Spoofed Akamai-Edgescape and X-Forwarded-For headers from untrusted peers are ignored by default.

3. Correlate branch offices, remote vantage points, and netperf​

Combining passive browser RUM with active edge probes answers whether slow page loads come from your CDN/origin or from a branch office WAN link.

A. Tag branch and site egress CIDRs (prefix_tags)​

  1. Open Settings -> Networks -> Prefix Tags (/settings/networks/prefix-tags).
  2. Add each branch or campus public egress CIDR with a consistent site tag (for example 198.51.100.0/24 -> site:branch-01, 203.0.113.0/24 -> site:hq-west).
  3. The Rust RUM collector matches every incoming beacon's client IP against prefix_tags in memory before IP truncation and stamps prefix_tag on the stored beacon. Click any prefix_tag row in Dashboards -> Web RUM to isolate that site's Core Web Vitals.

B. External L7 HTTP/TLS + MTR vantage points (no inbound ports)​

Rather than exposing bandwidth test ports to the public internet, deploy an outbound-only serviceradar-agent on an external VPS (such as an Akamai Linode instance in a remote region):

  1. Enroll the agent via Settings -> Agents -> Deploy so it dials outbound mTLS to your agent-gateway (50052/TCP) with zero inbound ports required.
  2. Open Solution Packs Catalog (/catalog) and install HTTP API & TLS Certificate Posture (io.serviceradar.http-api-and-tls-posture) targeting your website URL (api_target_url) and hostname (tls_target_host), assigned to that remote agent.
  3. Optionally attach an MTR Profile (/settings/networks/mtr) to track hop-by-hop internet path latency and packet loss alongside passive RUM.

C. Private WAN throughput & jitter (netperf over WireGuard)​

For site-to-site WAN links (such as a branch office or remote cluster linked over WireGuard or MPLS), use the embedded netperf runner inside serviceradar-agent over your private overlay:

  1. Open Settings -> Networks -> Netperf (/settings/networks/netperf).
  2. Create Netperf Responders on two enrolled agents using their private LAN or WireGuard addresses (port 5199/TCP mTLS + optional 5199/UDP HMAC -- never expose 5199 publicly).
  3. Create a bounded Netperf Profile (for example TCP bidirectional, 100 Mbps ceiling, 5s duration, 1s omit, 300s interval).
  4. Create a Netperf Policy (hub_mesh or hub_and_spoke) linking the profile and responders. Results stream into Dashboards -> Network Performance (/dashboards/network-performance), letting you compare site WAN throughput, loss, and jitter directly against prefix_tag-filtered LCP and TTFB in Web RUM.