Middle Eastern Exchange Points


it's flash heavy fyi

I'm talking to an ISP in eastern Nebraska who has a DS3 to Sprint. They've got a peer they tie to with private fiber. That peer has a DS3 from AT&T. Both normally see 20ms response times on pings from their border routers to the carrier router. Since last Thursday the Sprint connected ISP has been seeing this time bounce between 20ms and 1000ms or more. Customer experience with regards to latency mirrors the ping response times.

   Sprint has done intrusive testing on the DS3 physical layer/link layer - no problems found. The Sprint connected ISP has good bandwidth management practices - control of usage via an Allot Netenforcer and netflow export from their border router collected by a Manage Enginer Netflow Analyzer box. They see nothing out of the ordinary. I have less contact with the AT&T ISP but I'm told they've got some sort of netflow collection going and they've not seeing anything unusual in terms of customer traffic.

   Is anyone else seeing this? We're guessing there might be some shared MPLS transport for AT&T and Sprint in this mix since its affecting both ISPs in the same fashion, but we don't know the gritty details about the path. If this is an MPLS issue I believe Qwest transport could be involved.

Note that EMIX is a transit service, not really peering.

(It's peering in the same way that once upon a time the Australian for "peering" was "buy transit from Telstra".)