Perth was being served from Singapore the long way

Perth was being served from Singapore the long way

A logistics dashboard had users in Perth complaining of lag while Sydney users were fine. Same CDN, same continent. The edge node was in Singapore, geographically close, but the ISP's routing dragged Perth traffic east to Sydney, then north to Singapore, a 9,000 km detour for a 4,000 km hop.

The CDN's anycast wasn't the issue; the peering was. We worked with the provider to add a Perth point of presence and tuned the BGP announcements so the local ISP handed traffic off directly instead of backhauling it.

Perth TTFB dropped from 180 ms to 28 ms. The route went from 14 hops to 5. Dashboard interactions that felt sticky became instant, and the support tickets stopped.

The number that mattered: 180 ms to 28 ms, just by fixing the path.