Ob links

consonant (896763)
The Motherlode
The Cult


Please direct questions and comments (both yours and /.!) to the email address on the /. user page.

Listed on BlogShares


Thursday, April 26, 2007

Starts off alright, then...huh???

T1's (Score:4, Interesting)

by hackus (159037) Alter Relationship on 5:44 Thursday 26 April 2007 (#18878769)

Let me count the ways.....

1) T1, down at 3AM saturday. No problem, people are working on it.

Business class DSL: Bawahahahahaha...call who?
Cable: We don't work weekends.

2) Reliability. The infrastructure difference between DSL and Cable vs T1's are incredibly different.

T1's are simple in comparison compared to a DSL or Cable infrastructure. Too many people and too many things that can go wrong.

When I run a NAGIOS report for all my DSL lines and Cable lines and compare it to my T1 line over a complete 365 day interval:

My T1 had one incident in July of last summer with a direct lightening strike, was down for 3 hours. Didn't even have to call anyone, they had people working on it Sunday evening, and I got a voice mail it would be up in about 3-5 hours.

I have Sangoma cards in my Linux routers, and from what I can tell there was also a down/up event last year for about 1 second in my logs.

DSL line: It has had over 20 down events which I would call momentary lapses, 13 outright drops for 30-40 minutes at a time, and 80-100 quality alerts that indicated dropped packets or packet loss. I have two NAGIOS servers too, one for monitoring the internal network and one to monitor the outside network.

I made the NAGIOS box to monitor the ISP's so that I could tell if they were having an external or internal problem with thier networks.

6 times I had to dial in and remotely login to the AC strip and dump the power to the DSL unit to reset it, which would then "fix" whatever it was that made it loose its marbles.

One instance one of my facilities was down for almost 3 days, no DSL service. Something happened when SBC upgraded the line, as I asked SBC for a bandwidth increase. The SBC rep told me it was "standard practice" to change your IP address space with a line speed increase.

WT? When I pointed out changing the static IP's without telling your customers could have adverse affects on businesses VPN links, I got the "Well, thats what we do." I prompted told them to put the service I had back in place, they couldn't. They erased the passwords on the DSL modem and didn't have them.

They wanted me to drive 35 miles to a facility to put the password back into the modem.

I promptly dropped the DSL service. It didn't bother me anyway as all my locations have cable and dsl, linked through a BGP topology.

I also had the DSL modem replaced 8 times in the last 2 years at all 8 of my DSL/Cable facilities. The speedstream units suck arse. The netopia units are much better, but they still screw up once in awhile.

I even update the firmware myself, doesn't seem to make any difference so I stopped doing that.

Bottom Line: DSL saves money, it certainly does....but it isn't a 24x7 service, the customer service for business class sucks. For what you get with SBC business class cable its REALLY overpriced.

In fact, I would not call SBC business class cable anything remotely associated with "business". Its a consumer line with static IP's.

SBC can cackle all they want, but don't buy from them if your application needs anything but casual line use. It was so bad I had to buy cable as well so I could keep my facilities up 24x7.

This isn't limited to just one facility. I have Linux BGP routers in 10 facilities spread out over 50 miles. Every SBC facility equipped DSL service has the same issues.

Cable: Cable is better than DSL, only had 12 incidents. All of them related to the fact that the cable company keeps changing the signaling on the modems as the seasons go by. So, all 12 incidents were related to high packet loss due to bad signal. When they change the signaling to the cable modem, the line freaks, and they have to send a tech out to install a filter on the line. That must get REALLY expensive.

Cable is better, but running a BGP topology with multiple redundant pathways presented problems with cable and DSL.

For example, as our business grew over the years, redundant pathways was a must. We do logistics in a 24x7 context and downtime isn't an option.

So each of my 12 locations had DSL and Cable. It worked great for 2 years. When the cable would take a crap, BGP would route over the DSL and vice versa. It would take the router like 60 seconds to figure out the line dropped so automatically the BGP router would select the DSL route.

It was cheap to...

However, as I started deploying more and more Linux terminals/Tablets/desktops at all locations, I quickly found that bandwidth carrying capacity of DSL started getting to the point that a alternate DSL backup path for my cable modems basically became increasingly impractical. I mean I started pushing 320Kbits UP and although the cable modems could handle that, the DSL modems couldn't handle it.

So when I lost a cable path, network was SLOOOOOW.

So, now I have Norlight comming in and putting T1's into all locations.

I will be installing VoIP, so I need the bandwidth and the reliability, plus someone to fix the lines without me having to get out of bed at 3AM and call someone to alert them that their equipment was down.

My cable provider said he was going to miss me because my NAGIOS software was so much better than what he had, (i.e. HP openView) and that I became his canary for the fox valley when his network was sucking. "How will I know when my network is sucking?"

I told him to get a copy of Linux and start replacing his crappy Cisco gear.

Cable and DSL will be going away gradually over the next year. With it the pains of dealing with bgp pathways with screwy asynchronous upload and download speeds. T1 interconnects, with evenly matched sync up and down!!!

Wooooot! My network should be a gas! And of course I will be using all that fancy bandwidth to install 105 SNOM 320 phones with 3 sipxpbx linux boxes for the entire enterprise. 2 years later and Linux is EVERYWHERE.....even in the phones now.


Did I mention I run all of my facilities with openvpn and quagga/bgpd and kicked Cisco out on its arse?

I also of course, run iptables/tc to mark packets for QoS.

I have lots of legacy Windows Desktops, and they are my next target! They of course, must all die. :-)
Comments: Post a Comment