Carl Malamud: Internet Talk Radio, flame of the Internet.
This is Geek of the Week, and we’re talking to Geoff Huston, who’s technical manager for AARNET, the Australian Academic Research Network. Welcome to Geek of the Week, Geoff.
Geoff Huston: Thanks Carl, and g’day.
Malamud: So you’re from Australia. You have computers down there?
Huston: One or two. You know, we actually caught on to this trend a few years ago and and thought you know, this is something we should actually buy into.
Malamud: How big is the Australian Internet?
Huston: I think we’re now on a par with Germany, which considering we have about one quarter of their population isn’t bad going. We have about 75,000 hosts connected on the network right now.
Malamud: And how do you get into the Internet? What’s your way of connecting out to the outside world?
Huston: We run a satellite link across the Pacific and come in on the West Coast. At this point we come into NASA’s infrastructure on the West Coast and feed into the rest of the Internet across the US that way.
Malamud: So you come via NASA’s network. Now, NASA is a mission-oriented operation, the NASA Science Internet. They have fairly strict appropriate use policies. How does an entire continent become a client of NASA? And why aren’t you a client of NSF, or someone else?
Huston: Yes, getting a network started is always an interesting thing, and at some point always you carry the baggage of your initial efforts to get something together. Startup costs in Australian were really high, and the reason why they’re really high is of course that the little international link costs totally dominated everything. You know, when you’re coming across on a satellite bearer, in the Pacific satellite costs are much higher than transatlantic. There’s not that much business. So, the initial startup costs were extreme.
NASA in effect said, “We have a number of requirements within Australian of a general science infrastructure nature.” They were faced with the choice of running their own dedicated links with extreme cost, or simply saying, “Well if we can help you out with a US-based half-circuit initially, can you do the other side, the Australian half-circuit, and do domestic infrastructure?”
So it was a little bit of kickoff support that really started us going. And as long as we stayed within a general academic and research orientation, we have found the match with policies has been good enough to work on.
Malamud: How big is that pipe to the outside world?
Huston: Well this week it’s 768K. That’s a point of contention at this point because a country that size, we’ve used every last scrap of bandwidth on it. We’re going to a 1.5 meg link in August.
Malamud: In August, okay. And is that going to be enough? Is a T1 line enough bandwidth for an entire country?
Huston: No. When we installed the 768K link in March, we had three weeks of relatively clean service. But after three weeks the link was running at 768K all through the day.
Malamud: What does that do to service? Does that degrade catastrophically or is it just a gradual decline and things just get slower and slower and slower?
Huston: Well some things get slower and slower and slower. Some packets actually fight their way through quite dramatically. Telnet’s a really bad case of things getting quite slow. Odd enough, video and audio casts, those packets manager to fight their way through the morass quite successfully. So, the degradation is uneven, if you will. Some applications are worse than others.
Malamud: Australia makes fairly extensive use of video and audio. If you look at the multicast backbone you’ll see that there’s quite a few Australian sites that are on this worldwide multicasting virtual network. Is that an appropriate use of that scarce bandwidth out, and is there something you can do to control those users?
Huston: Well, the alternative is the plane system. And because of the way the airline market itself works, planes in and out of the country are also extremely expensive objects. Viewed like that, bringing the world to a desktop is a remarkably cheap way of getting things together. And the only penalty, if you will, of participating in things like an IETF or RIPE meetings and so on is loss of sleep for day or two. The costs involved in so doing are just phenomenally different. And as a country, and as a country that oriented technically and socially well towards North America and Europe, we certainly have a strong interplay with the technical activity in both of those areas. And increasingly I think the general science and research infrastructure. So we see this area as being one where quite frankly there’s an enormous implicit demand going on. And anything we can do to harden that to bring conferences back onto people’s desktops is going to make our service a service that’s really quite desirable in that country.
Malamud: So audio to video on networks is not just a high-tech toy. You actually are finding that is saving travel money or increasing awareness?
Huston: Well it’s saving travel money, but it’s bigger than that. Research these days is a collaborative effort that’s international and multidisciplinary. And you know, it might be okay for someone in America to hop on a plane to the East Coast on a fortnightly meeting, it might be okay to have meetings in Europe and America and get things done face to face. But when you look at a typical scientist or a typical academic inside Australia, one trip a year is considered to be…you know, you’ve reached the pinnacle of your profession at times. And the costs involved in travel and the time penalties, because it takes a couple of days to get over the time lags, well, all that says travel is not an option that realistically says we can involve thousands, tens of thousands of our talented people inside a productive area of research involvement. So from that point of view the Internet is…is critical to us. It’s not just a research project, it is a real strategy for a country. And we view it at that level of importance.
Malamud: Are you finding that your users and your funding agents also view that infrastructure the same way, or to them or are you just a research project?
Huston: I think they’re saying to us, “We’re watching you quite closely.” I think the point is that at this level, “Go hard. Justify your money. There’s no free ride in this. But if you actually manage to demonstrate you’re capable of providing us with real alternatives, and capable of giving us new opportunities nationally, then sure we’ buy it. But we’ll buy it on the basis that there are other ways of spending the money. You’ll have to justify your existence.”
So this is not a like a Gore Bill from the top. This isn’t a shower from heaven. This is quite frankly, “You do it right, you’ll get paid for your service. If you don’t, you won’t be relevant.”
Malamud: Do your customers pay or you or are you funded by some committee?
Huston: Well I maintain we’re actually a user-pays network. We’re funded by the constituents of the network, the universities, who act basically as underwriters. We gather money from connecting the network to people we deal with within the country—various people who want to take research, and various companies that supply us with goods and services. But in essence there’s very little government input, and there never has been. We’ve been forced to go through networking the hard way in saying, at every point along the line we’re gathering money from that user base.
Now, that user base has tightly-committed budgets. They’re saying, “We’re not gonna spend money on something else. We’re gonna spend it on you. You better be a better fit. You better be able to deliver more than where we were spending the money before.” If we can’t maintain that continual contract of quality, and that continual contract implicit, of relevant service, then we’re out of the picture. But so far I think we’ve been able to not only come to a competitive level, we’ve outstripped any other available means of doing this particular job. There’s just simply nothing else that offers us the capability, the range of service, at within orders of magnitude of the costs that we do it.
Malamud: Who are your users? Are the people the pay department chairmen in universities? Are they individual professors? Do you go outside of that community, outside of the academic world?
Huston: We do. We directly service some 360 organizations around Australia. Some thirty-eight of those are universities, of course. But of the rest we’re increasingly servicing relevant areas of federal and state governments. We’re increasingly servicing commercial entities that have research programs or participate in research, either the through its output or direct input at the start. And people who sell us goods and services. So, we’re using the network the way it should be used. I always believed that a network’s major power is ubiquity. That you’ve got to accommodated a diversity of motives, the same way as a telephone system, from figuring out what you’re gonna do next weekend to placing a country on a competitive position inside very strategic activities.
So, you’ve got to accommodate a broad range of motivations. The only way you do that is to make sure that everyone else is there. That what’s on your desk is better than a telephone and replaces it. And that’s the kind of attitude I think that’s going to see this becoming a productive and relatively high level of investment activity over the next four or five years.
Malamud: Geoff Huston, your network AARNET was founded fairly recently, right. It’s just a few years old?
Huston: Yes, we started this work in 1988. We’ve spent a lot of time inside that low-cost UUCP domain. And it was really a thing that was burnt by its own success. We got to the point where the telephone system and low-speed modems were incapable of delivering the backlog of mail until it got to the weekend. So that messages running from the East Coast to the West Coast were taking six days. We’ve found it very difficult to say well how can we improve it as a relayed-mail system. And if we’re gonna make a big investment we might as well shoot for something that was a service that affected far more people than a small community in mail at the time. So, in 1988 the decision was made to chase funding for an IP network.
And the timing I think was right at that point. Because until that time, it really did have that connotation if you will of a high-risk research project. But by the time in 1989 we were actually looking around it was a case of going to Kmart and looking at the range of routers on the shelves and saying well you know, “I’ll have the green one, thanks.” Going to the checkout, buying them and yeah, that was it.
Malamud: From the time you started your fundraising efforts to the time those routers were actually installed nationally was a remarkably short period of time. Is that because routers had become Kmart items at the time?
Huston: We spent nine months doing this work, and eight months three weeks and about fifteen days were spent making sure that the money algorithm was right. And around four days was spent on the router. You know, that kind of balance if you will, the technology was by then stable and readily purchasable. It was just off-the-shelf technology—which gave a lot of people warm fuzzies, too, because it smelt right. There was very little risk at that area.
So, from that point it was a quick effort. The actual installation of the network across the entire country was done across…sixty-odd sites, one way and another, inside three weeks.
Malamud: And you basically just got on airplanes and drove cars and just went from site to site, threw the router in and off you went?
Huston: Pretty much so. There were two of us at the time, and we spent four weeks without seeing each other. I went north and he went south, and we met back again in four weeks. And we did a lot of the sites. We actually didn’t connect them. Because of course routers come with three plugs. One plugs into the wall socket, and that’s pretty obvious because it sorta looks like a power socket. One plugs into the Ethernet and that’s pretty obvious because you can’t plug it into the wall. And the other one just simply plugged into the carrier’s modem. And again it was a big square plug and it was the only place it could go.
So typically by the time we got to the site, they’d turned it on. And of course we had arranged in the purchase to have them preconfigured inside flash memory. So that it was really a case of black box technology. They could plug it in, turn it on, and it was up by the time we got there.
Malamud: So that takes care of the routers. Now, do you find you have an easy time getting that bandwidth within Australia? Is a telephone company part of your operation? Are they one of your allies?
Huston: No, we are a straight commercial client. We actually haven’t got any deals at all with the telephone carriers. At this point we simply buy off the price book the same as any other client of the telecommunications company. In some ways it’s good.
In some ways what that says is we’re exposed to the true cost of networking from the start. We’re not inside a special policy regime because of some weird pricing structure. We’re actually facing a service at true cost. And the client base is there, again, at true cost.
Malamud: Is it true cost? Do you find that the pricing that you pay—you have two-megabit links on your national backbone, right?
Huston: That’s correct.
Malamud: Is the price you pay a fair price, or should it be less, should it be more?
Huston: Well, as an intermediary who buys and sells, if you will, then obviously it should be a lot lot less.
It’s a strange situation. I think at this particular point carrier pricing is based more on their marketing strategy than it is on a technology argument that says “This is the cost of the raw materials. This is the cost of our operation. This is what we can actually do in pricing.” But so often inside a small market—and Australia is in that kind of sense a small market—carrier operations tend to protect their existing investment. And generally view new technologies with a degree of suspicion. So that their marketing of say, frame relay in Australia, that doesn’t exist at this point. They’re not really interested in bringing in new product if it simply strips customers of all the product. Until it—
Malamud: X.25 being the older product in this case.
Huston: In this case X.25. They really want to protect that X.25 base for a while longer to get their money back out of the original investment. So, until customers actually walk right out of them and start doing private frame relay, I’m not sure the carrier would really move. It’s an artifact I think of a very small marketplace.
Malamud: Your network is routers and long-distance lines. Is that all AARNET does? Are you just a transit network or do you provide services? Do you have help desks?
Huston: We do a few things. We certainly do doorstop services to universities and the government research operations, so that we actually do the entire loop and the equipment on that side, where we’re dealing with core clients, if you will, of the academic and research support operation. For everyone else we provide a point of access. We don’t do an awful lot in terms of service, outside over the “Help, it’s not working. Let’s apply diagnostic tools. Let’s call in line technicians” and so on.
In terms of help desks and similar services we have said at this point to most of the university sites, quite frankly the applications, the services, the technologies haven’t changed because there’s a line going out the door. The same tools that go up and down the corridor, the same tools that go and provide library services to people’s desktops, are the same tools that negotiate the globe. It’s I think perhaps inappropriate to have a national carrier solving what realistically are internal issues about service and support. And we have argued strongly along the lines of saying, “You will get better participation on a site, and you will get more involvement and more support for your own investment in fiber-optic cables and routers and so on site.” Which is a much bigger investment than ours, by the way. You will get bigger payoff if you provide support that’s direct and local, and attuned to your own site’s characteristics.”
So from that point of view I’d argue not only havent’t we done an awful lot in terms of help desks and courses and so on, strategically perhaps it’s the wrong move.
Malamud: I could see with a large university where that would definitely hold true. And that they need to build their own local reliance, and probably have the ability to do that. But how do you handle kindergartens, and grade schools, and government offices that don’t have computers. Is that part of your mission to bring the Internet into those types of organizations?
Huston: At this point it’s not part of the mission of the network per se. But obviously those people who are involved in this activity, including myself, see a reason for our activities of saying we’re setting up a telephone system for the next century. And if you’re really looking at a goal that says this technology has a lot to offer; this technology is going to be widely-deployed; then yes, there is an argument that says we’ve actually got to carefully consider how to get individuals in areas that aren’t a rich infrastructure on to the network.
But I would claim rather than simply looking at help desks and support agencies, you’re really looking at a model it says, so far the Internet has relied upon connecting large-scale networks together. But there’s a new edge around there, and the edge is an access model. That rather than a large network, there’s a PC or a couple of of PCs. How do you support that? For 5‑year-olds, sure. For 18-year-olds, sure. For people at home. How do you actually integrate services onto that that give that individual, or that small workgroup, enough confidence that they can drive the technology? Enough confidence to feel that yes, there’s help out there if I need it. But at the same time say well, there’s no rules. It’s a very large network and there’s a massive opportunity for everyone, to gather knowledge, to work together and to communicate. And that model is something we need to work at globally. And it’s as I said, I think it’s more than help desks and training courses.
Malamud: Well what is it? It sounds like you’re saying you’re on your own at that point. That it’s a huge network and here it is, and have fun. Is that the way that we should be spreading the Internet? Relying on people to read the books and understand how it works? Or should we be spoon-feeding ’em? Should the interfaces be easier to use, for example.
Huston: When I go to the public library down the road, at times I just wander through stacks. And I don’t really know what I’m after. I’m looking for a cover that’s got a lotta colors. You know. Pick up a book and just simply browse it in a very serendipitous manner. And that kind of browsing is extremely exciting. You uncover things, you know, vague sort of references you’d heard in the past, and pick out something that catches one’s imagination. And I think is a real role for that.
There are other reasons why I go, associated with my work or whatever. And I would argue that in terms of structuring and environment on a screen, in terms of structuring a service, then realistically perhaps it’s service providers who wish to structure and spoon-feed rather than that network infrastructure itself. And like it’s not as if I deliver a telephone that says “I’m sorry,” you know, “just these ten numbers. From there you can go into other places, but originally only ten numbers and I’ve spoon-fed it for you.” You know, we actually have the telephone system in all its wonderful glory and depression, you know. It’s a telephone system. I think at the provider level, at a technology level, it is limitless. And it is, at this point, apparently boundless. We haven’t yet seen those edges. And to start creating artificial boundaries very early on…I don’t think perhaps is the right way of going about it.
Malamud: So it would not be right for you for example to decide that certain services are not appropriate on the network? Are you just a transit network in that sense?
Huston: Very much so. It’s often been argued, and we have I think as with any other network, gone through the debate of where the network operation starts and stops. I’ve consistently tried to put for the case of saying “Look, we switch packets, and our job is to really switch packets well.” After that, everyone else has got a job to do. And it’s a very valid job. It’s not for me to actually determine how people use the network. It’s up to them to use it and justify their own expenditure in so doing.
Malamud: Is your job different because you’re in Australia? Is your job as a transit network, as a national network, different than let’s say somebody in Europe doing a similar function?
Huston: There are levels that I think are similar but also there are things that are very ver different. Within Australia because of the very high cost of transiting the Pacific to get to anywhere else, the operation within Australia is much more of saying there really isn’t a lot of scale for 3,000 providers. It really is quite a valid case economically to say that the costs involved say if one provider does it openly, cleanly, and well, then it’s a remarkably cheap and competitive service to a country. And when I compare that within Europe, then there’s a valid case within Europe because of the higher density to say that different competitors come in and offer services. And that makes the role of the network provider one of interaction with other providers on a much more complex scale. We don’t have that kind of model within Australia. And perhaps I think that’s given us breathing space to offer a clearer vision of what a network can be to an entire country. And I suspect that if you divert that attention to say, [A/I S?] boundaries here and [A/I S?] boundaries there, sure it’s valid work if it’s necessary. But at times your key people are doing low layer work when perhaps they should be there saying it’s a very very big and strategic thing that we’re doing here.
Malamud: Well instead of lower-layer work, what are some of the upper-layer things you worry about? I noticed you have a big Archie server, for example. You have a service called Alex. Maybe you can explain what that is and why it makes sense in Australia.
Huston: Certainly. Within our network cost structure, at this point half of our expenditure on bandwidth goes across the Pacific. And the way our cost projections work, over two years around three-quarters of our network expenditure is going to be on going across the Pacific. That’s an extraordinary amount of money to try and access sites.
We did a quick analysis a couple of years ago and actually looked at where FTP traffic was heading and found that close to 30–40% of FTP traffic was just hitting ten sites. And from that point came the conclusion that if we bought a workstation and set up a three- or four-gig mirror, then for a fraction of the cost of one month’s rental, we’d have an extremely fast, local, accurate, cache. So from that point came the Archie project, which was a project to set up an Archie server but to actually also not only provide the pointers, but to point locally as much as possible, to bring in the most popular things.
What we found then, though, was to move on a bit to say, how can we cache? Well, obviously FTP is not a tool that implicitly lends itself to caching. Because normally the client goes directly to the server and pulls the file over. So what we said was, can we set up a file system on that particular machine that acts as if it’s the entire FTP archive of the world, in the same way as ASF or NSF offer wide-area file services, that map a directory hierarchy into the FTP world. And that’s the Alex file system that sits there. And in essence what it says is the first person to wander into Archie and pull a file using Alex actually does a transfer live. But obviously when a file gets announced, normally four or five people are going to pick it up inside a few minutes; and over the next week thirty, forty; and if it’s a popular one, thousands. If it’s sitting there it simply gets served locally. Presto, the amount of pressure on the international link gets reduced slightly. And from that point of view, given the costs involved in setting it up, it’s been quite a winner.
Malamud: You’re a trustee of the Internet Society, an international professional association on the Internet. Does this group help you in Australia? Do we need an Internet Society to help integrate Australia into the global community?
Huston: I think we do, but I think that the reasons why people look towards such a body vary from continent to continent and region to region. I sense within Australia an Internet Society provides some degree of ability to participate. In essence it’s not a case of being a relaxed consumer—or a very anxious consumer of someone else’s technology and someone else’s developmental model. It actually says to Australia, to a country that has extremely high aspirations on technology development, “We welcome your participation.” And at that level, it’s an initiative which I think is an extremely productive one from out perspective.
If the Internet effort was sitting quite squarely within the United States political boundaries, and if the funding continued to be a strategic investment dominated by the US federal government, then obviously we would find our ability to participate somewhat lower. And in that sense, gaining a national commitment, saying to governments, saying to commercial bodies, saying to the market at large, “This technology is really something that offers you ten‑, maybe even twenty-year strategies.” Without that ability to participate, without seeing the Internet as a technology removed from the initial players is a bigger thing, then one can say that honestly, that yes you can participate; yes this is something that if you wish you can invest in it too.
Malamud: Well how do we get people in Australia participating? Obviously you can. You’ve got a travel budget. But how do we get the young engineer to become part of this world? Does that person fly to the INET conferences once a year? Do they fly out to the IETFs in Amsterdam or in the United States? Or— How do we get these people integrated into that world community?
Huston: We mentioned before about audio and video services on the network, and quite frankly because within a country like Australia travel budgets begin at the middle or end of one’s career rather than at the start, the video and audio tool is going to be the way that we’re going to ignite the passions and the motivations of graduates and graduate students, I believe. Within Australia those such people will need a track record, if you will, of participation before they can put their hand up for funding. Now, we’re working very hard on a model that was similar to Europe around three or four years ago of trying to in some way set up a local area of engineering, and local activity, and say well, if you participate in that at your cost, then we will try and get you to meetings at our cost as a network provider.
But at this stage such thinking is quite early on. And I would really say that if you asked today what the tool is, then the tool is I can bring the IETF to your desktop. That’s not hard. So, you can participate. And you should participate. If you take it from there, then if you take on more commitments, if you start doing working group chairs, then your argument for support from the host institution where they’re employed and their argument of support from the network provider, becomes far solider. And their argument for strategy, if you will, of saying, “We’re bringing technology back into the country to be developed,” becomes a very clean one.
Malamud: Well thank you very much Geoff Huston. This has been Geek of the Week and we’ve been talking to Geoff Huston from AARNET in Australia.
Huston: Thanks, Carl.
This is Internet Talk Radio, flame of the Internet. You’ve been listening to Geek of the Week. You may copy this program to any medium and change the encoding, but may not alter the data or sell the contents. To purchase an audio cassette of this program, send mail to radio@ora.com.
Support for Geek of the Week comes from Sun Microsystems. Sun, The Network is the Computer. Support for Geek of the Week also comes from O’Reilly & Associates, publishers of the Global Network Navigator, your online hypertext magazine. for more information send email to info@gnn.com. Network connectivity for the Internet Multicasting Service is provided by MFS DataNet and by UUNET Technologies.
Executive producer for Geek of the Week is Martin Lucas. Production Manager is James Roland. Rick Dunbar and Curtis Generous are the sysadmins. This is Carl Malamud for the Internet Multicasting Service, town crier to the global village.