- NANOG - lists.nanog.org

MAE clarifications: Full Duplex FDDI, Congestion
by feldman＠MFSDatanet.COM 09 Dec '96

09 Dec '96

There's been a flurry of information about the MAEs over the weekend, and I'd like to clear up a few points: Full Duplex FDDI The DEC Gigaswitch supports full duplex FDDI by default, and MFS has *not* disabled this feature. At least one provider has already plugged in a Full Duplex capable interface and seen it work without any intervention on our part. There's no additional charge for use of full duplex FDDI, and there's nothing anyone has to do to "sign up" for it. I would ask that you notify me so I can watch more closely for difficulties, however. Cisco's full duplex FDDI interface is a module for the VIP2 card, which I believe is only supported on the 75xx and 70xx/RSP (but not 70xx "classic") platforms. I don't know of any other router vendors supporting full duplex FDDI, but they don't always check with me first! MAE Congestion There have been problems at both MAE East and West recently with apparent packet loss on trunks between Gigaswitches. In both cases, the links themselves were not full, but the packets were being dropped on ingress to one of the Gigaswitches. The apparent reason is head-of-line blocking in that switch (Giga1 at MAE East, the Ames switch at MAE West) caused by one or more providers filling their FDDI access links. In both cases, we have been working with those providers to find ways to relieve pressure on their attachment ports, and have made some configuration changes to reduce the amount of packet drops. We are also working with DEC to find solutions for the head-of-line problems. Steve Feldman MFS

1 0

Re: Building exchanges that matter ..
by Vadim Antonov 09 Dec '96

09 Dec '96

Joe Rhett <joe(a)Navigist.Com> wrote: [Gb ethernet] >...and any technology which experiences performance problems at 40% >utilization will is questionable at best... Last time i used switched ethernet it didn't experience any problems at 100% load. But you just can't do ATM w/o a switch. Talk about flexibility. --vadim

2 1

Re: Building exchanges that matter ..
by Joe Rhett 09 Dec '96

09 Dec '96

> Very fast ethernets are designed with two modes, standard > a standard CSMA/CD bus and a full-duplex collision-free mode. [ .. your lesson on basic networking deleted .. ] > ATM since one is stuck playing with forwarding based on > only one address space. Yah, I'm familiar with all this. I'm sitting behind one of those "Bad Notworks" boxes right now, actually. > I'm not sure what flexibility you're referring to wrt ATM > vs very fast ethernets, however the latter in combination > with software in one particular modern router that isn't > completely rocket-science can spoof rate-limited VCs > based on MAC addresses, which more-or-less duplicates > the one feature of ATM that I happen to like. Uh.. I'm not sure what you're asking, because I suspect the answer I'm about to give you'd have already considered, no? Ethernet -> Flat. Creating true architectural redundancy isn't possible without separate routes for each connection. Also, even full duplex Ethernet never achieves full bandwidth utilization due to the protocol implementation (hardware protocol). Advantage: All vendors support it, and is usually works without help. ATM -> Flat/Star/Web/whatever. You can create both bandwidth and redundancy without routing. Disadvantage: Every vendor supports it differently, and most of them only have 2 people that really know it. Never works without help. Also, congestion control between hetrogenous systems is a no-op.

3 2

Re: Exchanges that matter...
by Rob Liebschutz 09 Dec '96

09 Dec '96

> On Sun, 8 Dec 1996, Joe Rhett wrote: > > > > > > It is actually quite hard in practice to push a FDDI switch > > > into trouble, although theoretically it should be simple, modulo > > > the interesting effects that token-withholding can produce. > > > > Hm. Not sure what you mean. FDDI can certainly be over-utilized, that's > > not hard. On the other hard, FDDI doesn't fall down and die like > > Ethernet does when congested. > > > > > In practice, the limiting factor will tend to be the per-port > > > bandwidth rather than the aggregate bandwidth. One would have > > At the Atlanta-NAP we offer full duplex FDDI, why not try to get MFS to do > it? Cisco now has a full duplex FDDI card, so you can do 200 Mbs into the > NAP. Talk to Steve Feldman from MFS. At the October Nanog meeting he mentioned that their Gigaswitches (not sure if this is was just for MAE West) were ready to support Full Duplex FDDI and he invited volunteers to try it. Unfortunately, the GRF does not yet have a full duplex FDDI interface available, but when I asked about it, Ascend told me that they had implemented a CDDI interface in 3 weeks time to gain the business of of one customer. Rob

2 1

Re: Exchanges that matter...
by Vadim Antonov 07 Dec '96

07 Dec '96

Joe Rhett <joe(a)Navigist.Com> wrote: >The issue that everyone seems to be overlooking is network design >flexibility. Even if ATM has higher overhead, you can only grow a FDDI >network so much. It really isn't all that hard to max out most FDDI >switches. Nah. The issue is that ATM is a rather expensive way to achieve nothing. And, if you never heard about it before: FDDI is not what people use to send data over backbone lines, no matter what trade mags show in their cute pictures. --vadim

1 0

Re: Exchanges that matter...
by Vadim Antonov 07 Dec '96

07 Dec '96

salo(a)msc.edu (Tim Salo) wrote: > From: Paul A Vixie <paul(a)vix.com> > [...] > ... and they aren't subject to ATM's cell tax ... >I am surprised, (well, maybe not), that you aren't concerned about >the excessive overhead present in FDDI networks... FDDI is not a WAN technology. Losing 20% of bandwidth of a 10 ft piece of fiber is one thing. Losing 20% of bandwidth of a $3M/yr circuit is quite different. ATM as a LAN is probably ok, if it can compete with Gb Ethernet price/performance-wise. --vadim

2 1

Re: Exchanges that matter...
by Scott Huddle 07 Dec '96

07 Dec '96

"Alex.Bligh" <amb(a)xara.net> wrote > > I think that what Alex really should say 'No'. What he saying is > > that for Xara UK-sourced traffic is split like this: > > 50% is sent to UK peers at the LINX in London > > 40% goes to the USA > > 10% goes to dGIX in Stockholm > > > Doh! Yes. Must get reading lessons. Missed the word "remaining". So none of the traffic from Xara goes back out to Xara customers?! > Currently I have no foolproof way to differentiate traffic that > goes over the LINX destined for in UK and outside UK. There's > some work going on to build a list of UK ASes to this end (amongst > others) but it hasn't been done yet. Quite agreed this is hard stuff, thus my interest :) > Jim is quite right (except it works bidirectionally). Do you see symmetrical flows in and out of the country or do you see more inbound that outbound? -scott

3 2

RE: Exchanges that matter...
by Danny Stroud 07 Dec '96

07 Dec '96

Yes, thanks for making my point. des ---------- From: owner-nanog(a)merit.edu on behalf of Vadim Antonov Sent: Thursday, December 05, 1996 3:59 PM To: Danny Stroud Cc: nanog(a)merit.edu Subject: RE: Exchanges that matter... Danny Stroud <dannystroud(a)msn.com> wrote: >So how do you reconcile the increasing number of private facilities? Although >not exchange points in the truest sense ala MAEs and NAPs, they do carry an >increasing percentage of cross-network traffic. The number of these types of >facilities will grow. Maybe our difference here is that I count these and >others do not. des Er. Private exchanges are there only because public exchanges are running out of gas. They have the same problems with routing as public exchanges do. --vadim

1 0

SLC Connectivity
by Daryn D. Fisher 06 Dec '96

06 Dec '96

I know this isn't the correct forum for this question so please reply directly to oz(a)tpa.net instead of the list. We need another T1 connection in Salt Lacke City, UT and really can't find anyone we want to use out here. Any suggestions? If you are trying to sell your own service please submit estimated pricing and service guarentees. thanx

1 0

Re: Exchanges that matter...
by Dave O'Shea 06 Dec '96

06 Dec '96

> > Do you see symmetrical flows in and out of the country or do you > > see more inbound that outbound? > > There is a third possibility, of course. In our case we see more > outbound traffic towards the USA then inbound traffic from the USA, > because our UK customers have Web sites popular in the USA. .. So we have not only a trade deficit, but an "ack" deficit to keep politicians awake at night.. :-)

1 0