This is a networking question at the top tier level.
Why does ATT, and others, deliberately insert a long delay (500 ms) in all packet traffic at major gateways. This is plain obvious for traffic across major links (Pacific, or to Asia, etc). I can ping plott over a few hours, and it's easy to see this is deliberately inserted delays, and manipulated as they choose.
Is this the major players trying to channel traffic through other major connnections? Or them giving priority (or non priority) to various major ISP's as they choose? Or them pretending they have run out of band width?
Yes I know at this time the Taiwan problem exists, but this happens all the time year round, and fluctuates at very predictable amounts at various times of day.
Generating ICMP responses is somewhat CPU intensive on a large scale, and the router CPU is used for other important activities like maintaining routing protocol peering. There are a couple different approaches to avoiding resource starvation due to ICMP. First off, all vendors treat ICMP with low priority by default -- the router will service all other CPU tasks before handling ICMP. After that, it's a matter of throttling; Juniper takes the approach of throttling ICMP after certain CPU thresholds are exceeded to preserve the integrity of the system. Cisco, on the other hand, doesn't wait for the CPU to start to climb and begins rate limiting responses even on an idle processor. Your response time on Cisco backbone nodes is going to vary quite a bit depending on how many other people are hitting them with ICMP at the same time.
Enterprise Network Engineer :: Hosting Hobbyist :: Master of Procrastination
"The really cool thing about facts is they remain true regardless of who states them."