- Test - lists.nanog.org

Re: Fractal models of Big-I Internet
by Vern Paxson 10 Oct '96

10 Oct '96

> Given that real world systems that exhibit fractal behaviors can now often > be modelled mathematically, do you know of anyone who has attempted to > apply such fractal models to Internet traffic? There have been several studies presenting strong evidence that Internet traffic in general, and also that of some individual protocols (of particular interest: HTTP), have a fractal structure. This is still a major research area, though. One of the important unanswered questions is whether the fractal structure is a good model for behavior on short time scales (10-100 msec). Best guess (IMHO :-) is that it isn't. So there's a major question of how to put together a model that deals with both the short-term and long-term dynamics, along with major questions as to what's the *right* fractal model. Unfortunately, the math gets hairy fast, and the Internet is tremendously diverse. > When designing protocols > do researchers take this fractal nature into account? This also is a research frontier. Some work's been done though. One study that comes to mind is that by Sugih Jamin of USC (now at U Michigan) and colleagues, on the impact of fractal traffic on admission control schemes. I don't know of any protocols being designed with fractal traffic in mind. I'm right now in the middle of analyzing a bunch of end-to-end data to try to build up some sort of model of congestion time scales. One thing I'll be looking for is a link between the measurements and fractal models. I'm hoping to present some results at the first 1997 NANOG. Vern

1 0

SNMP recording and browsing tool
by Mark A. Lovell 10 Oct '96

10 Oct '96

I'd like to announce the release of a new network analysis tool, the SNMP Vulture (http://www.vix.com/vulture) Vulture is a set of utilities to collect and store SNMP data and browse it in an interactive fashion (using any web browser). The result is a tool that can give you a good grasp of what a large network is doing. Here's a bit of the main man page: VULTURE(1) VULTURE VULTURE(1) NAME vulture - collect a round of snmp based on configuration files SYNOPSIS vulture [-dhv] DESCRIPTION Vulture collects SNMP data and stores it in a hierarchy of log files. It determines what data to collect based on a configuration file that in turn relies on a series of SNMP template files to determine which objects need to be col- lected for each entry in the configuration file. When invoked, vulture will collect a single set of data and store it in the log files. This is not very useful, and vulture is almost never invoked this way. This is usu- ally only necessary while debugging changes to configura- tion or template files. Vulture collects its data in a two pass operation where it first sizes up the tables involved, and then requests entire tables of relevant information. This is an effi- cient process, since it knows which objects it needs out of given tables, based on the information in it's template files. All data that can be directly matched against the net- work.desc file are placed into the regular log files. Data for interfaces that are not listed in the network.desc file are placed in a parallel set of discard log files. These data are usually discarded much earlier than regular data.

2 1

RE: Internet II is coming...
by Sean Doran 10 Oct '96

10 Oct '96

| "IP Unter Alles", in other words, right? No, not at all. IP under computer-to-computer data. WRT your list, I don't think that either videoconferencing or connectivity to the PSTN is a particularly interesting thing to do with the Internet, in that I don't see either as being a fundamentally important Internet application. Both are features, and some kind of voice and video with what humans consider acceptable lag is kinda neat, particularly when considering things like sex-on-demand, but this is not what the Internet should be designed for, since it's not likely to be the majority traffic by revenue attractor. So, I think that having a parallel infrastructure for real-time traffic is reasonable right now, and my business sense tells me that such an infrastructure would be much more ripe for metered-use than any combination network, whether built on ATM or IP or both. MBONE people will tell you about the fact that videoconferencing works quite well on various IP internets and the Internet now, despite the fact that the MBONE is still fairly primitive in terms of a global multicast plant. CUSEEME works reasonably on cheap equipment and dialup lines, and seems to scale aesthetic response against money fairly well, despite the fact that it's also a fairly simple first-cut. Various people, including a Sprint Canada reseller, will tell you that ALGs exist that convert various voice-over-Internet things into signals that can be dealt with by PBXes and other telephony equipment. There are at least two SS7<->IP gateways in development with what appears to be short times-to-market ahead. One of these is being built along the lines of what PNO like an RBOC or a PTT or a cable company might want to deploy in their own facilities. These are neat hacks which will do fun things to international and national long-distance tariffs worldwide, in that you may suddenly have two QOSes for voice: one which is a side-effect of IP connectivity and which is of iffy-to-tolerable-to-good quality, and one which has metered, often distance-and-time-of-day- sensitive charging and a generally predicatably very-good-to-excellent quality. So, wrt your point 4 (connectivity to PSTN needs ATM), you are right that you can't do it just with Cisco equipment, but you can't do much else just with Cisco equipment, as I don't think they build SONET/SDH MUXes, CSU/DSUs, phone switches and the like anyway. Neither do the currently-popular ATM vendors, for that matter. Go figure. We're now down to 155Mbps medium speed to the desktop. I imagine that we can avoid arguing about the small details of cell tax and the like, and start looking at Gbps ethernet and other forthcoming LAN issues. They really are LAN issues; the Internet vs global ATM really isn't a factor here, nor is really IP vs ATM on a protocol-to-protocol basis. | Oh, and by the way, given that the local loop provider has OC-48 SONET | provisioned to this particular location, we could just as easily have | provisioned the connection to our backbone at OC12 as opposed to OC3. Did | I miss the Cisco announcement of an OC12 IP-SONET card? You may wish to discuss an NDA presentation on the forthcoming generation of routers from each of Cisco, Juniper and Bay Networks. Nothing on the router market now really handles OC12 as a medium rate, whether we're talking POS or ATM or anything else. OC12 is too slow anyway, frankly. | [ATM] | does have some undeniable advantages, and one of them is its ability to | carry voice traffic independent of IP traffic, and to connect in a | reasonably straightforward manner to the PSTN. Would you prefer to have | voice traffic clogging up the IP backbone?? Well, the infrastructure telcos use now, whether based on SONET or on plesiosynchronous stuff, separates PSTN and IP and X25 and Frame Relay and even ATM and so on and so forth. I expect this to continue. The only argument in favour of using ATM rather than TDMish stuff really comes from the mystical promise of ABR, which, so far, doesn't seem to work well when mixing certain traffic loads together, notably enormously-aggregated TCP and anything bursty. If, however, a sizeable amount of dead space really could be recovered from various parallel networks, thus freeing up more bandwidth along fibre paths, I'm all for that. The problem is that this doesn't really work well just yet, and I'm disinclined to believe that it's likely to work well with big-I Internet traffic ever, thanks to its rather fractal nature. So, just to clarify things, I don't discount ATM as a technology entirely, I just think that its time as something between IP and SONET for big-I Internet applications (other than *maybe* for customer aggregation and the like), is ticking away... Sean.

2 1

Re: Internet II is coming...
by Vadim Antonov 10 Oct '96

10 Oct '96

Deepak Jain <deepak(a)jain.com> wrote: >I suspect that they will be going completely switched and have very few >total points on the whole network (depending on what configuration they >go to). A network of OC12s or OC48s in a redundant star will have >significant performance benefits because 1) no routing, or very symmetric >routing. Except for what is in the center. Or is piling trouble into one place makes less of it? I guess just opposite. >2) very low latency <8ms coast to coast I'd suspect. Sure, dude. Faster than light networks are coming. FYI, latency in rouetrs in the existing backbones is only about 15% of wire dalays. >3) priority queues, quality of service, reserved bandwidth, etc. And nice salaries to the activists and "research" grants for something any qualified network engineer knows as A-B-C. >I am not even touching the Mae-East at 30% fantasy. All I know is that the >UUNet <--> Sprint OC3 private connect at Tysons Corner is at better than >24Mbits average and mostly limited to router CPU problems. FYI, that interconnect _is_ a part of planned extension of MAE-East which is conviniently located next to Tysons Corner (Tysons Corner is a shopping mall). The ablility to bypass shared medium by simply dragging a 10' piece of cable is why the colocated IXP architecture was selected in the first place. --vadim

1 0

RE: Internet II is coming...
by Jim Browning 09 Oct '96

09 Oct '96

>From: Sean Doran[SMTP:smd@cesium.clock.org] >Sent: Wednesday, October 09, 1996 2:28 PM > >Finally, I assert that we are already at the point where anything >that can be deployed today that is based on ATM (which incidentally >typically rides over SONET/SDH) can be kludged up (or even done right) >with Cisco gear and SONET/SDH. ATM has a temporary edge in being >multivendor and making it easy to do TDM-style and point-to-multipoint >things, however the former is likely to be short-term and the latter >is something that can be done better for Internet traffic anyway, >with a bit of cleverness in the latest and in the next generation >of IP routers. "IP Unter Alles", in other words, right? Okay, I'll take the bait. This past weekend, we demonstrated (1) video-conferencing, (2) streaming video, (3) 155 Mbps to the desktop Internet connectivity, and (4) connectivity to the Public Switched Telephone Network, all over our ATM OC3c backbone, simultaneously and continuously. #1 & #2 can be either IP or native ATM, and in this case we ran it over IP. #3 requires ATM, although you can get close to it using other technologies. But what about #4? Connection to the PSTN? Please explain to me how to, in a non-kludgey manner, provide the PSTN connectivity using only Cisco routers. I certainly believe that most customers would consider speaking into a microphone attached to their PC and listening to the PCs speaker just a tad on the kludgey side. Not to mention that they would need an extra sound card to avoid having to say "over" every time they were through speaking. No, I think they want to use the nifty 900Mhz Sony cordless phone they just bought at Circuit City, and have a level of service (note I avoided the dreaded 'Q' word) similar to what they get from their POTS line. Oh, and by the way, given that the local loop provider has OC-48 SONET provisioned to this particular location, we could just as easily have provisioned the connection to our backbone at OC12 as opposed to OC3. Did I miss the Cisco announcement of an OC12 IP-SONET card? >Since these routers will be needed with or without ATM, the time >to ponder whether ATM really has that much added value in the >long run is upon people already. As is the time to ponder how all the various technologies can work well together. What is long past is the time to flame a technology which has already become an integral part of the Internet infrastructure, and which will remain so for a long time to come. There are plusses and minuses to the widespread deployment of ATM, just as with *any* technology, however it does have some undeniable advantages, and one of them is its ability to carry voice traffic independent of IP traffic, and to connect in a reasonably straightforward manner to the PSTN. Would you prefer to have voice traffic clogging up the IP backbone?? -- Jim Browning "My views do in fact reflect the views of my sponsor, and I am now going to go look for that asbestos suit said sponsor provides to those of us who are inclined to support the merits of ATM on Nanog"

2 1

Re: Internet II is coming...
by Sean Doran 09 Oct '96

09 Oct '96

Come on, Vadim, this is a great idea. They should build this cheaper faster better network to be used exclusively for meritorious traffic because it's in the national interests of the USA. Moreover, the the members of the Internet II consortium (hi University MIS types and management!) and any governmental funding body involved should mandate that any meritorious traffic use IPv6 *only* and set an end-to-end IPv6 option to indicate which traffic should go across Internet II and which should go across whatever IPv6 Internet exists as provided by Evil Commercial Interests at the ridiculously high prices which are being charged universities and research labs. Moreover, they should also mandate that Internet II be built on ATM, use ABR, and also use RSVP to guarantee the kinds of qualities of service that the R&E community requires. This would be a boon to network researchers throughout the USA. You can bet that I (not a U.S. national, and an Evil Greedy Commercial Bastard) also will be singing the praises of the folks from PSU, Stanford and Chicago, and their EDUCOM ally Mike Roberts and their representative George Strawn, the NSF's chief proponent of Internet II-like initiatives, because it's a fantastic idea, and the implication that it is either pork or a bad or unworkable idea is one that is beneath you, Vadim, even if the proposers leave the other requirements out of the final solicitation, award, or both. I mean, you can't possibly believe that the industry will solve or even wants to solve the issues most of the folks at University campuses have been complaining about, just as you shouldn't believe that it can't use a push to deploy ATM, ABR, RSVP and IPv6. A group of some thirty-four or so high-power customers is exactly the sort of push that will finally just get this stuff done. I can't wait until other countries jump on the bandwagon. Maybe the EU will ressurect its similar ideas, or the G7 will keep going with their talks, or maybe this could end up right in the lap of the U.N. That would be way cool. Sean.

10 12

Re: Internet II is coming...
by Sean Doran 09 Oct '96

09 Oct '96

| I suspect that they will be going completely switched and have very few | total points on the whole network (depending on what configuration they | go to). A network of OC12s or OC48s in a redundant star will have | significant performance benefits because 1) no routing, or very symmetric | routing. 2) very low latency <8ms coast to coast I'd suspect. 3) priority | queues, quality of service, reserved bandwidth, etc. #0: well, nobody's physical topology is shaped like a redundant star. at very fast rates, for (2) above in particular, having the network topology follow rights-of-way is a good plan #1: *something* somewhere has to do routing, and has to know the full details of what the topology looks like and make decisions based on that information and any constraints. Problem A in the current Internet is that boxes that have to know everything are feeling the crunch of complex topology and instability; I don't see how Internet-II can get around this, and moreover, I do see how easily additional parallel infrastructure can worsen Problem A for everyone (and in particular for the Internet-II connectees). Also, have you noticed Vern Paxson and Van Jacobson saying that path symmetry really doesn't seem to matter with respect to TCP throughput, or that path asymmetries even on point-to-point circuits has existed for some time in the real world? I dunno, maybe very symmetric routing might help, but aren't you building some bad failure modes by trying to force this by building a star? #2: Are you planning on building photon accelerators? BTW, what kind of delay do you get between Stanford and UCSD if you have a star topology? You can assume that your physical topology is built out of the four principal right-of-way carriers' fabric, whether leased or otherwise. #3: What type of per-packet and per-flow time budgets do you think you have at OC12 and OC48? If you can make all this happen with existing technology (or can make new technology), then there are at least four companies that will want your resume ASAP, particularly if you can do this and accomplish your #2 (assuming you accidentally missed a digit) simultaneously. Sean.

1 0

Re: Internet II is coming...
by swolff＠cisco.com 09 Oct '96

09 Oct '96

>Even with "recreational" use of the Internet, the Internet today is much >faster then it was 6 years ago (T1 national backbone (singular)). <ahem> UUNET had one. PSI had one. NASA had one and so did DoE. Oh yes NSF had one too. Stephen Wolff Business Development ciscoSystems 202 337 1213

2 1

RE: Internet II is coming...
by Sean Doran 09 Oct '96

09 Oct '96

| The institution in the situation you describe, procure an ATM interface to the | provider. Have the provider two PVC's. Set both PVC's for EPD. Contract | the Researchers PVC as a VBR PVC, and contract the other as a ABR or UBR | PVC. In other words, somewhere close to the edge of a network fabric, some router is making a decision based on things like origin/destination or protocol number or what-have-you to send traffic across a point-to-point line instead of the Internet. Instead of building a real point-to-point line, you build a PVC using ATM, and then play games to make sure the ATM fabric does the right thing or some approximation thereof if there is contention in the ATM fabric. In the brave new world of an ATM-free future (assuming the ATM people don't get _cheap_ SVCs out the door before we get the "Peter-Tony Protocol" model deployed), you can do the same for less cost by pulling a tributary out of your SONET MUX at the appropriate bandwidth, and doing policy-based routing using your favourite vendor's NetFlow switching or some future tag-switching-like system or the equivalent. This is your PVC, no EPD or PPD needed. Cells? What are those? All each approach is doing is source-routing across different paths, making a decision somewhere close to one or both edges of a given fabric. This could as easily be accomplished by having a fabric made up of either multiprotocol routers which deal with IPv4 and something else which does source-routing, or of routers which have the time budgets to do Christmas-tree packets (all decorated with options). If you make decisions to do different queue/drop algorithms based on something decided at the edge of a fabric, or at a host, then you might not even necessarily need any sort of separate infrastructure to meet most reasonable QOS-like demands. This is your *BR game-playing with ATM, for dealing with contention. The separate circuit model has an attraction in that you can avoid having Special Traffic move normal traffic out of the way, and you avoid paying a huge penalty in routers in the core of the best-efforts Internet, both of which are serious problems in efforts like RSVP. A mix of separate circuit/parallel infrastructure vs. smart queuing in IP routers should really be done using the appropriate parts: real circuits and real routers. Playing games with ABR and friends introduces the problems of disconnected resource consumption feedback loops (TCP vs ABR, for example), which tend to compete with each other in nasty ways. Temporary circuits of whatever flavour have an attraction in that you can get a mix of parallel infrastructure and smarter infrastructure without necessarily incurring enormous cost, and without worrying about byzantine billing for normal Internet traffic. This is coming, both for ATM and for non-ATM models. Finally, I assert that we are already at the point where anything that can be deployed today that is based on ATM (which incidentally typically rides over SONET/SDH) can be kludged up (or even done right) with Cisco gear and SONET/SDH. ATM has a temporary edge in being multivendor and making it easy to do TDM-style and point-to-multipoint things, however the former is likely to be short-term and the latter is something that can be done better for Internet traffic anyway, with a bit of cleverness in the latest and in the next generation of IP routers. Since these routers will be needed with or without ATM, the time to ponder whether ATM really has that much added value in the long run is upon people already. Sean.

1 0

Re: Internet II is coming...
by Kent W. England 09 Oct '96

09 Oct '96

At 11:28 PM 10/8/96 -0400, you wrote: > >About the broader issue of the Internet II project, I have always >suspected that the "retail" market (100 M households) and the R&E >market (1000 largest R&E and industrial institutions) had incompatible >network requirements. The consequence is that no infrastructure could >be optimal for both. It would be a *good*thing* if the providers >could develop two parallel infrastructures The fiber infrastructure and the multiplexer infrastructure are good enough to support both Internet and Internet II. What the Internet lacks for Internet II today is IP tunnels with QoS. So I think that one IP infrastructure could be good enough for both and will be some day. This I believe. :-) >Internet II and the like are good things. No question! > >--MM-- > > If 100k university students, faculty and staff can demonstrate large scale personal videoconferencing, that would be a big win and a good market driver. BTW, is someone going to build traffic monitors and diagnostic tools into the Internet II infrastructure? Any new inter-domain routing protocols? Source demand routing? QoS? Or are the universities satisfied with just more bandwidth? --Kent

1 0