| gabriel rosenkoetter on 11 Dec 2003 19:37:02 -0500 |
|
On Thu, Dec 11, 2003 at 05:57:08PM -0500, Andy Smith wrote: > I have read with interest your reports of problematic connectivity using > your DCANet DSL service. My name is Andy Smith and I'm the senior > engineer at DCANet. I've also been lurking around PLUG for quite a > while now. Thanks for taking the time to reply, Andy. I'm still a Speakeasy customer because I still expect to move out of your region but stay in their's, but I still recommend your service for folks who are in and intend to stay in the Philadelphia area, and in no small part because you pay attention to this mailing list. > Your DSL circuit is a Verizon ADSL circuit, which means Verizon provides > the loop, and they carry your traffic back over their ATM network to > DCANet's facility in Philadelphia. Really. So the ATM link isn't actually COVAD equipment? (Or are you not a COVAD reseller? I thought you were...) > Sometimes in this whole process, and lately more frequently, Verizon > will re-provision which ATM PVC your particular traffic traverses. This > happens with no warning to us or to anyone else. That's insane. I'll wager they notify their internal IP services when they do this. (Uh, well, actually, based on their quality of service, maybe not.) How is this not monopolistic behavior? Does your contract with them really say they can do that? (If so, yowch.) When I settled for ADSL (with Speakeasy for IP and COVAD for DSL), I never had this problem. But that was two years ago now. > The problem is that our router has a route for your particular > IP address pointed to the specific ATM PVC that your ADSL modem > lives on. Couldn't the customer's modem be configured to arp for the addresses they're supposed to have? Will the customers *outgoing* packets still be routed in this case? And replies to those packets? (What the original poster seemed to be describing was just a loss of addressing, not a loss of connectivity from the inside.) > Usually the end user just has to release / renew DHCP (or, reboot, > which does the same thing). When the computer comes back up and > requests the IP address via DHCP, our router knows to install the > /32 route pointed to the ATM PVC from which the DHCP request arrived. > All is well. Surely you can see how, for those of us using operating systems where the typical problem fix is NOT "reboot", this isn't much of a solution, nor is it much of an "always on" connection. That said, I'm pretty sure that your routing would *also* get fixed when the customer automatically renewed at lease expiry, so passing out leases not too much longer than Verizon's mean-time-to-screw- everyone-else-involved-for-not-giving-them-money, you should be able to fix this magically without the customer having to do their own releasing and renewing. (I'm presuming that your network and DHCP server could cope with the load, especially since any sane DHCP server implementation will randomize the lease times a bit to keep itself from being hit with a DoS.) > Even though the IP address is assigned with DHCP, we still abide by our > long standing policy / feature of assigning static IP to end users. So > you'll get the same IP address every time from the DHCP server. Incidentally, I think you're doing this the Right Way. > I checked my router config, and your IP addresses have static routes > defined. One way we could try and troubleshoot this for you would be to > remove these static routes, and have you switch to using DHCP for IP > acquisition. And if the original poster (sorry, I've deleted the other messages in this thread already) would like some help on how to make his firewall and other configuration cope with getting a DHCP address, I'm sure he could just ask here. (Getting a DHCP address doesn't mean it's going to change, inherently. I'd much rather that all the statically-assigned IP addresses in my network at work were actually DHCP-managed. It makes so many things so much easier...) > This will solve the potential problem of Verizon changing > the ATM network on us, and will hopefully eliminate this as a possible > source of trouble. If by "solve" you mean "the poster has to renew his leases when he can't get to his system from the outside world to tell it to renew his leases". If you actually *are* setting your lease times based on how frequently Verizon's been breaking the rest of the world, my apologies for the tone of that. > If you need a circuit for critical services, or something that is more > stable and more reliable, we can certainly talk to you about loops from > Covad, AT&T, Worldcom, Verizon T1, etc. The prices on traditional T1 > circuits have really dropped in the last few years, to the point where > they may be acceptable to certain users. I think that may be my cue to point out that my Speakeasy-managed COVAD SDSL (768) line has had troubles a grand total of zero times in nearly two years of service. Note that I pay a little bit over three times as much as DCANet's ADSL subscribers do for this service. It makes the biggest threat to eclipsed.net's being up be PECO (and, let me tell you, PECO is *flitty*), but APC sells some nice ways to solve that problem. Considering I'd want broadband at home anyway, this is significantly less expensive that colocation (or, more to the point, it's that level of networking service in my home, where I can blame noone but myself for colo-faciliy issues like AC and power). > We're all adults here. Well, in fairness, no we're not. In fact, we've had a PLUG meeting presented by a pair of highschool juniors in the not-to-distant past. :^> That said, we usually try to *act* like we're all adults... -- gabriel rosenkoetter gr@eclipsed.net Attachment:
pgpBaQrixz2Az.pgp
|
|