June 2000 - Test - lists.nanog.org

Re: PMTU-D: remember, your load balancer is broken
by Paul Vixie 19 Jun '00

19 Jun '00

akyol(a)akyol.org ("Bora Akyol") writes: > FYI, there are core routers out there that can support line rate > HW-based fragmentation. I believe this is necessary in any router > with heterogeneous interfaces. OK, somebody check me on this. "Line rate" means back to back, minimum spaced packets. Fragmentation means turning single packets into multiple packets. Each such packet will have its own headers, thus, fragmentation means turning a single header into multiple headers. Headers take space. Therefore if there's a 100Mb/s FDDI (MTU 4400) on the input side of a router, and a 100Mb/s FastEthernet (MTU 1500) on the output side of that router, then there WILL be lossage due to fragmentation, since "line rate" on the output side has more IP headers octets per IP payload octet than "line rate" on the input side did. This is a contrived example but there are more complex and more subtle flows in real life that would excite this problem. Fragmentation is like classful addressing -- an interesting early architectural error that shows how much experimentation was going on while IP was being designed. -- Paul Vixie <vixie(a)mibh.net> SVP for Internet Services, MFNX M.I.B.H. Inc. is a subsidiary of Metromedia Fiber Network, Inc.

1 0

Re: Private and Public Peering
by Paul Vixie 19 Jun '00

19 Jun '00

At 21:42 06/18/2000 -0700, Steve Feldman wrote: > The closest to a real distinction that I've been able to come up with is > whether or not a third party is involved *and* whether packets (or cells) > are switched. smarcus(a)genuity.com (Scott Marcus) writes: > What most people mean by "private peering" is a direct interconnection > between two providers. That's most often implemented over a circuit > between the two, without either deploying equipment to the other's > premises; again, however, that's simply a matter of engineering > convenience. These connections are conceptually point-to-point. I like Steve's distinction better. When folks peer at PAIX, they do so with a "direct interconnection between two providers". The fact that both providers have extended their networks to PAIX, or are using PAIX as a POP, doesn't make the "peering" less "private". -- Paul Vixie <vixie(a)mibh.net> SVP for Internet Services, MFNX M.I.B.H. Inc. is a subsidiary of Metromedia Fiber Network, Inc.

2 1

Re: Pacbell Internet NOC #?
by Walter Prue 19 Jun '00

19 Jun '00

Dan, >Does Pacbell Internet have a NOC phone #? You might try 800-463-8724. It is possible that that 800 number only works in California. I don't know. Walt

1 0

Pacbell Internet NOC #?
by Dan Hollis 19 Jun '00

19 Jun '00

Does Pacbell Internet have a NOC phone #? Jared's NOC list doesnt list one, only an email address. -Dan

1 0

Re: Jumbo Frames (was Re: MAE-EAST Moving? from Tysons corner to reston VA. )
by Richard A. Steenbergen 19 Jun '00

19 Jun '00

> On Mon, 19 June 2000, "Bora Akyol" wrote: > > > As long as most end users are running Ethernet, Fast Ethernet, DSL or Cable > > Modems, what is the point of jumbo frames/packets other than transferring > > BGP tables really fast. Did any one look into how many packets are moved > > through an OC-48 in 1 seconds. (approx. 6 million 40 byte packets). I think > > even without jumbo frames, this bandwidth will saturate most CPUs. > > > > Jumbo frames are pointless until most of the Internet end users switch to a > > jumbo frame based media. > > > > Yes, they look cool on the feature list (we support it as well). Yes they > > are marginally more efficient than 1500 byte MTUs ( 40/1500 vs 40/9000). But > > in reality, 99% or more of the traffic out there is less than 1500 bytes. In > > terms of packet counts, last time I looked at one, 50% of the packets were > > around 40 byte packets (ACKs) with another 40% or so at approx 576 bytes or > > so. > > > > What is the big, clear advantage of supporting jumbo frames? > > When 1500 byte frames from the customer's LAN enter the customer's > router and enter some form of IP tunnel, then a core fabric which > supports larger than 1500 byte frames will not cause fragmentation. > It's not necessary to do the full jumbo size frames. I suspect that > supporting two levels of encapsulation will be enough in 99.9% of the > cases. For the sake of argument, what would be the downside of using a > 2000 byte MTU as the minimum MTU in your core? When the next end user upgrade is deployed and everyone has devices which can support larger MTUs, wouldn't it be a shame if they said "if only the internet core ran at larger MTUs, we could negiotate higher MTUs and make everyone happier". Also, it is far more then "marginally more efficient". For every packet you deal with, there is a great amount of work doing routing lookups, dealing with memory management, and handling interrupts. Copying another few bytes of data is easy in comparison. Since we are asking GigE to act in a server and backbone role, we should acknowledge that the requirements will be different from the average end-user ethernet. One of those requirements is that the backbone should be able to pass larger packets it may encounter without resorting to fragmentation (which only gets harder and we start getting into higher speeds). Aside from that, and the fact that there is nothing harmful in supporting larger packets through your network, there is the fact that if we want people to support standards we KNOW are good for them (even if they don't), we have to actually ask for it. Imagine an internet with a reliable MTU negiotation mechinism, which can take advantage of improved thruput, much lower CPU usage, zero copy, page-flipping, DMA transfers, and all those other lovely things. These are important for many reasons. Without these techniques, we can't even do line rate GigE on "common place" servers, let alone have any CPU left over to do more then just send packets. Its easy to just say "we'll throw a server farm at it" or "we'll just get a faster processor", but as higher speed links become more common place, and as GigE becomes common in servers (when servers can actually use it effectively) and 10GigE becomes commonplace for backbone links, we'll start to see these things matter. Why engineer ourselves into a corner of shortsightedness which only gets harder and harder to fix, because its "easier" to do nothing? (sorry Michael, just using your msg as a good point to reply :P) -- Richard A Steenbergen <ras(a)e-gerbil.net> http://www.e-gerbil.net/humble PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)

4 3

Where in the world is ?
by Sean Donelan 19 Jun '00

19 Jun '00

> Sean Donelan was not heard from, and we're all hoping the power gnomes > didn't get him :) California should be safe this week, I'm staying at CNN Center in Atlanta. Of course, if CNN goes off the air, you'll know why :-) Is the CNN tour worth the 45 minutes? With brewery tours, at least you get a sample of the product at the end. In unrelated news, the State Department lost all its phones on Thursday including phones in its operation center. An earlier fire damaged its backup batteries, and when rain knocked out utility power, no phones.

3 2

Private and Public Peering
by HANSEN CHAN 19 Jun '00

19 Jun '00

Hi, Can someone tell me the difference between private and public peering? A peering paper I have read explained the switch is shared in the case of shared and non-shared in the case of private. I still do not understand. Appreciate if someone could shed some light on this matter for me. My guess is that public peering is done in a commerical internet exchange while private peering is done in a facility owned by either one of the ISP. Thanks, Hansen

3 2

Re: MAE-EAST Moving? from Tysons corner to reston VA.
by michael.dillon＠gtsip.net 19 Jun '00

19 Jun '00

On Sat, 17 June 2000, "Richard A. Steenbergen" wrote: > > What is the largest MTU that you can use across a GigE fabric? > > I think this would be extremely foolhardy. The common number that all > those using jumbo frames should support is 9000 bytes (not 9k aka 9216). > This would probably need to be a condition of using GigE at the NAP, > in order to achieve any kind of common support, and you're eliminating > some vendor gige support (assuming their network can even carry the 9k > packets TO the nap without requiring even more work and possible > problems). I really should have worded that question more carefully. What I meant was, what is the maximum MTU that is supported by all GigE manufacturers and thus can be guaranteed to work without fragging across any arbitrary GigE fabric? I was under the impression that the answer is 1500 because not all vendors support larger MTUs. And the implication is that GigE (in general) is not suitable for a public exchange fabric because it will introduce fragging in packets which are encapsulated (tunnelled). Of course, it would be possible to choose GigE equipment carefully and to configure it correctly to avoid this downside. But it seems to me that it is essential for any exchange point fabric or private peering link or IP backbone to support MTUs larger than 1500 edge to edge. --- Michael Dillon Phone: +44 (20) 7769 8489 Mobile: +44 (79) 7099 2658 Director of Product Engineering, GTS IP Services 151 Shaftesbury Ave. London WC2H 8AL UK

1 0

Re: Jumbo Frames (was Re: MAE-EAST Moving? from Tysons corner to reston VA. )
by Vadim Antonov 19 Jun '00

19 Jun '00

Bora Akyol <akyol(a)akyol.org> wrote: > What is the big, clear advantage of supporting jumbo frames? There are none, not without doing something about TCP congestion control. The bigger frames are, the more coarse the congestion control is, and the flows are short enough on average to make it barely effectual. Increasing frame size to 10kb makes most flows "mice" flows, which kills the slow-start algorithm. Of course, flows may start by sending smaller frames first, but, again, what's the point? With 1500 byte frames, the framing overhead is about 2.8%, however, given that the 40% of packets are only 40-41 bytes long, this seems to be a very minor effect. --vadim

1 0

Re: PMTU-D: remember, your load balancer is broken
by Paul Vixie 19 Jun '00

19 Jun '00

Valdis.Kletnieks(a)vt.edu writes: > Has this changed? Has "fragmentation" become a Great Evil, ... ? Yes. http://research.compaq.com/wrl/techreports/abstracts/87.3.html says: Research Report 87/3, December 1987 87.3 -- Fragmentation Considered Harmful Christopher A. Kent, Jeffrey C. Mogul PostScript Version gzipped PostScript Version Adobe Acrobat PDF Version Internetworks can be built from many different kinds of networks, with varying limits on maximum packet size. Throughput is usually maximized when the largest possible packet is sent; unfortunately, some routes can carry only very small packets. The IP protocol allows a gateway to fragment a packet if it is too large to be transmitted. Fragmentation is at best a necessary evil; it can lead to poor performance or complete communication failure. There are a variety of ways to reduce the likelihood of fragmentation; some can be incorporated into existing IP implementations without changes in protocol specifications. Others require new protocols, or modifications to existing protocols. wrl-techops(a)pa.dec.com Copyright © 2000 Compaq Computer Corporation I was there, I saw the research that went into this, I know the guys who did the work, and I agree completely with the conclusions thus presented.

3 2