Following. Good thread.
Location doesn't matter with anycast CDN
tomchan
Member
paul76 said:
Origin still drowns when the cache misses
This is why I run origin in Vultr London with autoscaling rules on CPU. Cache miss spike = origin scales in 90 seconds. Not instant, but survives.
Without that? Yeah, origin dies. CDN doesn't fix capacity, only latency.
works on my bench ¯\_(ツ)_/¯
tomchan said:
CDN doesn't fix capacity
Finally someone says it. Anycast is a latency hack. It is not capacity. Your origin still needs to handle the long tail of misses, purge requests, and dynamic junk.
Location matters for capacity planning because capacity costs differently by region. See Carl's power numbers.
Has anyone tried running origin on Contabo and CDN on Cloudflare? Heard Contabo's bandwidth is actually capped in practice despite "unmetered" claims.
Andrew73 said:
Heard Contabo's bandwidth is actually capped in practice despite "unmetered" claims.
Not capped, but shaped. Their Singapore routing is Cogent-heavy to some eyeballs. You get the bits but not the path quality. CDN masks this by pulling over better transit, but you're still at mercy of Contabo's upstream mix — https://bgp.he.net shows the AS-path story.
Location dead? Only if you ignore AS-path.
visit twice: install and decom