Since Sandy went through, Comcast is having issues communicating with Verizon. If you have Comcast, you may have trouble getting to the blog. Comcast and Verizon peer with each other through a third party called Tata Communications, which has its data center in North Jersey, which of course is without power and suffering in the aftermath of a hurricane.
We are on a 25Mbit/25Mbit Verizon FiOS connection here with the blog. My employer is on Comcast, so that means I have been unable to work from home, and have been in the office. When in the office, I have either no communication or unreliable communication to home, so blogging suffers. We will return to a more regular schedule once things get back to normal.
Seems to be working now!
I’m on comcast and I’m seeing the blog. I could see it at work too, and we’re comcast there.
It’ll probably depend how you’re routed. Not everyone on Comcast will route through that data center, but a lot of folks in the area will.
I tracert though Lancaster then to ashburn, VA.
Yeah, the handoff in Ashburn to Verizon is what’s handled through Tata. It seems to be working OK now. I’m going to head home and see if it’s working from there too.
Interestingly I traceroute from Lancaster, to Pittsburgh, to NY, then to Verizon.
If only we had some routing protocol that would automatically update when peers went down…. :)
There are issues with that. E.g. every non-leaf node router has to maintain a map of where to send a packet to get closer to its destination. Rebuilding that map after a route goes down is … expensive. Worse, the overall system of maps has been observed to behave in very strange ways under these sorts of loads. I haven’t followed this sort of thing for a very long time, suffice it to say it’s very non-trivial.
There’s also the difference between peering and transit. Big providers like Comcast and Verizon agree to peer, generally for free, with each other, since it keeps down the load on their routes. To fix the mentioned problem, one or both of these companies would have to start providing “transit” as well. I.e. packets enter their network and then travel a long way before reaching their destination. Not counting hardware failure situations like this, transit always costs more money and as far as I know therefore always comes with a charge.
I.e. you don’t “peer” with someone else such that you carry traffic to and from their customers long distances across your network, i.e. I couldn’t set up a facility in one location and then ask providers to peer with me.
To end, you would normally expect problems like this to be solved quickly, i.e. route around the damage, but we can be very sure the network engineers at Verizon and Comcast are insanely busy getting their networks back in operation and dealing with the issues that each step of this presents.
On the other side of the Sandy Zone here in Greater Cleveland, Time-Warner/RoadRunner is acting really Gonzo. Up, Down, sometimes it comes in, but won’t let you go out, etc.
I’m SOOOO Glad that my stuff is “Stand Alone” and not “Bundled.” Too Many Eggs in One Basket still holds true.
Agreed on the eggs principle, but as for your connection’s flakiness I wonder how much of it is due to your using a cableco instead of a telco that at least once upon a time aspired to true five nines uptime. (And they really were serious, I worked with some of these types from the Cincinnati Lucent group on a media gateway project (the only thing with a dual IETF and ITU standard) and the requirements and restrictions are amazing.)
Compare to cable companies who I’m told in the bad old days would stick pins in a map as they got calls of service outages until it was clear which repeater had failed.
I keep my parents and myself on AT&T DSL; it isn’t fast or cheap and the caps are a bit low (more than a bit if you’re into video), but it’s rock solid.