- NANOG - lists.nanog.org

RE: Faster 'Net growth rate raises fears about routers
by Roeland Meyer 04 Apr '01

04 Apr '01

> From: Travis Pugh [mailto:tpugh@shore.net] > Sent: Tuesday, April 03, 2001 10:34 AM > I'm at a multi-POP network in Boston. We've had great luck selling > customers a Verizon circuit into one of our POPs and a > Worldcom circuit > into a different one. It costs more, but they don't have nearly the > exposure of a single circuit customer. However, if you're > not set up to > do this, the appropriate level of paranoia calls for circuits to two > different providers. Maybe if SPs really addressed availability > requirements of their customers, it wouldn't be such an issue. The problem with this, if done, is that we back right into the other problem of prefix filtering. If the customer has a /19 or /20, there is generally no problem. But, if it is the usual case (/24) then only one of the upstreams can aggragate the routes up. What is the other ISP to do? How would this be made to work? BTW, this is exactly the reason we weren't fully multi-homed yet. Yes, greg described a way where both interfaces (end point) were NAT'd. However, I have a concern with brittleness and tinker-factor there.

3 2

Need Help
by doug＠safeport.com 04 Apr '01

04 Apr '01

Your network center is really closed?? I do not have my customer number with me. You can lookup by company name: Safeport Network Services. My first question is when will you be back in the office? _____ Douglas Denault doug(a)safeport.com Voice: 301-469-8766 Fax: 301-469-0601

2 1

Provider change (was RE: Faster 'Net growth rate raises fears about routers)
by Sean Donelan 04 Apr '01

04 Apr '01

I'm sure the telesales people are making those phone calls as this thread goes on. The FDIC has this down to a science with banks. The bank closes on Friday evening, the FDIC swoops in, and by Monday all the accounts are moved to a new bank. FEDWIRE reroutes all the traffic to the new service provider, and the checks "in-float" don't even bounce. Besides the potential route table explosion as the network is deagregated, another problem is how do you verify who has what authority to change which circuit id, ip address, etc. Does everyone just start grabbing address blocks, do providers include the CIDR block as a line item on the invoice. On Tue, 03 April 2001, Randy Bush wrote: > what if an isp with a lot of dedicated line customers in old A space takes > the dive and a few thousand of their customers come to isps each asking to > have 38.42.666.0/24 or whatever routed? > > mirjam says > o route them now, forcing instant renumbering would add insult to injury > o give them the normal 90 days to renumber > > and she's never led me wrong so far.

1 0

RE: Faster 'Net growth rate raises fears about routers
by Roeland Meyer 03 Apr '01

03 Apr '01

Sorry, I thought we were discussing all multi-homing. Your example doesn't help the business whose ISP suffers a business failure (such as DSLnetworks, Flashpoint, et al), only the case where the access provider fails SLA. To put it bluntly, a single circuit, to a single business, while annoying, doesn't cause wide-spread outages when it fails. It is thus, the lesser-order case. The case I thought was under discussion is when an ISP dumps something on the order of 10^3 or more customers when they fail. I understand that NorthPoint abandoned ~100,000 customers when they sold their backbone to AT&T and AT&T didn't pick up the subscribers. I will wager that many of them were /24s. DSLnetworks had over 700 Covad customers, FlashPoint was larger. For various definitions of "wide-spread", this is a much larger issue than a broken copper-pair. I suspect that it also has a much higher likelyhood of occurance. Especially, in the current business shakeout. Guess what ... it won't stop. This sort of problem will be with us forever. We should find a solution ... someday ... ya think? Business failures are on one side of the problem and CIDR aggregation is on the other. > -----Original Message----- > From: RJ Atkinson [mailto:rja@inet.org] > Sent: Tuesday, April 03, 2001 11:26 AM > To: Roeland Meyer > Cc: nanog(a)merit.edu > Subject: RE: Faster 'Net growth rate raises fears about routers > > > At 13:55 03/04/01, Roeland Meyer wrote: > > >The problem with this, if done, is that we back right into > the other problem > >of prefix filtering. > > No. These are separate tail circuits to separate POPs > of the same ISP. So that one ISP only needs to advertise its > fully aggregated prefix. So the problem you postulate does > not arise in this particular situation. > > > What is the other ISP to do? > > You didn't read closely enough. 2 tail circuits, > 2 POPs, but only 1 ISP was the scenario outlined. It works > quite well, provided one picks the ISP thoughtfully. > > Ran > >

4 3

RE: Faster 'Net growth rate raises fears about routers
by Roeland Meyer 03 Apr '01

03 Apr '01

> From: Bill Woodcock [mailto:woody@zocalo.net] > Sent: Monday, April 02, 2001 2:51 PM > > On Mon, 2 Apr 2001, Roeland Meyer wrote: > > From what I just went through (DSLnetworks v Covad), > this is straight > > bulldust. Had I been properly multi-homed, MHSC would > NOT have been down for > > 5 weeks. > > Patrik said ISPs, not IAPs. So, please explain to me how not being multi-homed is anything other than a bad-thing and high-risk? No, I am not including colo, because it is assumed that you know what their arrangements are before you "buy". Reputable colos are multi-homed, in spades.

9 9

RE: Faster 'Net growth rate raises fears about routers
by smd＠clock.org 03 Apr '01

03 Apr '01

randy, that's called "moral hazard". Sean.

1 0

RE: Faster 'Net growth rate raises fears about routers
by Richard A. Steenbergen 03 Apr '01

03 Apr '01

On Mon, Apr 02, 2001 at 03:37:22PM -0700, Bill Woodcock wrote: > > > So, please explain to me how not being multi-homed is anything other than a > > bad-thing and high-risk? No, I am not including colo, because it is assumed > > that you know what their arrangements are before you "buy". Reputable colos > > are multi-homed, in spades. > > You say "responsible cab drivers must have not one, but two taxicabs, > in order to provide service in the event of a failure. Therefore, I > bought one from Fisher-Price, and one from Hot Wheels, and I'm > astounded to find that neither provides me with the luxury which I > expected." I think Patrik may have been suggesting that if you had a > Checker, you might not need to worry quite so much about redundancy. Only one transit? For a reliable internet, two transits is the minimium requirement, and you have but one, which is less then two, and two is what you need... Curses... You must immediately purchase some transit, which you need for internet, for without transit you cannot have the internet that you so require. -- Richard A Steenbergen <ras(a)e-gerbil.net> http://www.e-gerbil.net/ras PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)

5 4

RE: Faster 'Net growth rate raises fears about routers
by Roeland Meyer 03 Apr '01

03 Apr '01

> From: Eric A. Hall [mailto:ehall@ehsco.com] > Sent: Tuesday, April 03, 2001 12:10 PM > Convincing customers that it is cheaper/better to put their > main servers > somewhere off-site away from them is the challenge. Otherwise > more of them > would do it. You understate the problem. Service bureaus and ASPs have one problem in common, MIS/IT departments are loath to place their corporate data (jewels) under the control of another. Convincing them otherwise is an up-hill battle that you will mostly lose. To a lesser extent, this is also true of colos. The other part of the argument, that the local (fully burdened) square-footage is almost always cheaper than decent colo space ($1,200.00 per 1/4 rack, per month, typical). You therefore have two strong motives for locating corp datacenters on-site, rather than off-site. Either one, by themselves, are difficult to overcome. Then you have the third argument; most corporate data centers are used more internally than externally (road-warriors and B2B ops not withstanding). For performance reasons, you want the servers closest to their users. Unless you can get 100baseTX to the off-site facility, this means a strong incentive to mount servers at HQ, not in the colo. We are now back to multi-homing a business. BTW, this entire mess was played out before, in the 70's and 80's ... when the service bureau biz died. At that time SNA APPN/APPC was the network (my briefs were dyed "blue" at the time). The end-game was proprietary pre-internet data centers, connected by mutiple slow links.

1 0

RE: Faster 'Net growth rate raises fears about routers
by Irwin Lazar 03 Apr '01

03 Apr '01

> <quote> > "Half of the companies that are multihomed should have gotten > better service > from their providers," says Patrik Faltstrom, a Cisco > engineer and co-chair > of the IETF's Applications Area. "ISPs haven't done a good enough job > explaining to their customers that they don't need to multihome." > </quote> > > From what I just went through (DSLnetworks v Covad), this is straight > bulldust. Had I been properly multi-homed, MHSC would NOT > have been down for > 5 weeks. Good point, one of our customers multihomed to two different providers and had several instances where a BGP configuration change by one of their providers made their networks invisible to half of that provider's peers. These types of errors take quite a while to track down and are generally only found when the help desk gets calls that some people from around the world can't get to internal network resources (Looking glass sites are invaluable in these situations). Risk mitigation from these types of errors is easy justification for multi-homing. Irwin

2 1

RE: Faster 'Net growth rate raises fears about routers
by Barrows, Jeff 03 Apr '01

03 Apr '01

It is perhaps useful to note that there is at least one network out there with over 210,000 network entries in their tables today-- with multiple routers in that same AS having well over 100 BGP sessions on them-- comprised of around ~25% iBGP/eiBGP sessions in peer groups, and the remainder being eBGP sessions. These boxes push a lot of traffic, deal with a large/flat IGP, do IP filtering, and a variety of other random services. That same network had ~182,000 network entries in their tables in October of 2000. This could be reduced, and the rate of growth slowed, if necessary, but there apparently hasn't been a compelling enough reason for them to do so yet-- which doesn't necessarily indicate anything beyond the fact that there hasn't been enough pain in that area for them to change anything yet. Not to say that great pain isn't on the horizon. - jsb I don't doubt that this will continue to be a topic that is over-sensationalized, as is done here: http://www.nwfusion.com/news/2001/0402routing.html Which includes this tasty quote: "By the time the table gets to around 200,000 entries, we may be pushing a default-free router well beyond its processing capability." ...and this one: "Half of the companies that are multihomed should have gotten better service from their providers" [...] "ISPs haven't done a good enough job explaining to their customers that they don't need to multihome." > Granted, but are you saying that the 15 minutes it takes one of my BGP > sessions to reload has no relevance to the crusty, old processor doing > route calculations on 104,000 routes? Multiply the CPU > available by 5, and then look for bottlenecks. Seems sane to me. > Also seems a hell of a lot easier than trying to redesign the network > characteristics in 1 year and implement them ... IPv6 anyone? > > -travis > > > > > *shakes head* people keep forgetting this. Do you guys also > > think you can solve the internets problems by adding more bandwidth? > > > > > > > > Adrian > > > > -- > > Adrian Chadd "The fact you can download a > 100 megabyte file > > <adrian(a)creative.net.au> from half way around the world > should be viewed > > as an accident and not a right." > > -- Adrian Chadd and > Bill Fumerola > > > > > >

1 0