AT&T Midwest → Re: Routing Issues in Chicago

said by flashtro:

---

I'm also experiencing reduced Speeds.

899 downstream
245 upstream

on the 1.5Mb package.

---

LOL.....Here is my advice to you then:

"The issue seems in the throughput of your line. There may be errors that are being generated on your line or other factors that is causing the lower download speed. Have Beachboy or one of the other Ameritech/Asi guys run a test on your line."

Btw, been there, done that. Once about a FEW more people report it, then maybe someone will look into it..... may have already, as my throughput seems back to "normal" (for right now).....LOL.

Thanks for the advice, doesn't apply here though. Maybe it does for your line?



EDIT: clarification

[text was edited by author 2003-08-03 19:59:11]

Just because your experiencing throughput problems doesn't mean there is a routing problem. There may be congestion at one of SBC's external peer connections but that doesn't mean that the routing is not working. The BGP routing protocol does not change next hop destinations for routes when there is link congestion or latency for any reason. Only if the peer session goes down or if a metric or AS hop count changes in the routing table will the route point to an alternate destination.

You mentioned there was a lot of latency between hops 6 & 7 on your traceroute.

6 10 ms 10 ms 20 ms asn174-cogent.eqchil.sbcglobal.net [151.164.248.
42]
7 50 ms 60 ms 60 ms 154.54.1.18

That is an external connection from SBC to Cogent down at the Equant facility. If that link is saturated it isn't necessarily SBC's fault. Typically a service provider will maintain an average of 65% link utilization or less to any particular provider and upgrade as necessary. If there happened to be a burst of traffic at some point time passing through that link that caused latency/bandwidth issues, it sucks but that is how the Internet works. SBC will pretty much agree that it sucks, but get a helmet. Unless they are regularly bursting to congestion points on that link, they won't upgrade it.

said by rolande:

---

That is an external connection from SBC to Cogent down at the Equant facility. If that link is

---

Equant?
Try Equinix

I consider congestion at any point on a line = routing problem. If someone doesn't fix/redirect the traffic (ie, "re-route"), the congestion grows and continues - This is where we disagree.

At least I see the "bursting" (to congestion) was affecting more (users) than just me.....

I know that re-routing does work (around the "bursting to congestion"), as when this happened before (also on a Sunday) a few weeks ago, a tech re-routed the bottleneck, and it was all done..... that's all folks....that's my 2cents.

said by DeeC:

---

I consider congestion at any point on a line = routing problem. If someone doesn't fix/redirect the traffic (ie, "re-route"), the congestion grows and continues - This is where we disagree.

At least I see the "bursting" (to congestion) was affecting more (users) than just me.....

I know that re-routing does work (around the "bursting to congestion"), as when this happened before (also on a Sunday) a few weeks ago, a tech re-routed the bottleneck, and it was all done..... that's all folks....that's my 2cents.

---

The problem is in Cogent, it's not a "routing problem" but either:

Capacity issue on Cogents side
Something wrong with the circuit between SBC & Cogent

Routing problem is when your packets don't make it to the right destination, or go the *really* wrong way. This is a standard case where you have something wrong between peers. These things typically get resolved somewhat rapidly.

ok, whatever you want to "call" it...its still a recurring problem, and lets hope it gets fixed....

debating over semantics (the wording of it) doesn't resolve the actual issue(s) as they each arise...

thx for the clarification though!

Equant Internet Exchange = Equinix Same difference. I have a buddy who is a Field Engineer for Alcatel and has done several SONET deployments at that facility and they refer to it as the Equant Midwest Hub.

Yeah, if you had a routing problem you'd be more than pissed because you wouldn't be getting ANY throughput. If SBC is continually having capacity issues on certain peer links, it sounds like they are not maintaining the 35% breathing room for peak traffic loads.

Big service providers should monitor their aggregate traffic flow through a common set of POPs that share egress bandwidth and build those peer links to support a 65% load during peak hours. That leaves 35% for growth and lead time to increase capacity and still support the rare extreme peaks of traffic that may occur through certain connections. The problem in this case is you are dealing with DSL. When you pay next to nothing for cheap commodity bandwidth you get a backbone that will perform like a cheap commodity. SBC isn't earning revenues for that bandwidth to warrant spending much money to build out their peer connections with a 65/35 ratio.

said by rolande:

---

Equant Internet Exchange = Equinix Same difference. I have a buddy who is a Field Engineer for Alcatel and has done several SONET deployments at that facility and they refer to it as the Equant Midwest Hub.

---

Is this the same place on Cermak (right near the McCormick place)?

Yes. Quite a few carriers have big SONET POPs in that building. My buddy just finished another contract job there for France Telecom. They turned up 4 OC-192's each to Nashville, Cincinnati and Denver, passing through Level 3's POP on Canal St. too.