Juniper Networks Master of IT, Paul Schopis: OARnet

February 25th, 2007 |
Image for FaceBook
Download Audio FileRight click here to download audio | Share The Connected Social Media Player

 

Eighty-eight colleges and universities. Two-and-a-half million downstream users. Paul Schopis is associate director of OARnet, based at The Ohio State University, and as such he oversees an expansive network for users with a wide range of needs. In this podcast Paul discusses the challenges he faces in building and maintaining such a network, shares some knowledge learned in his ten years at OARnet, and explains the fish problem.

This Juniper Networks podcast is part of the Juniper Networks Master of IT program.

Transcript:

Host: Paul Lancour – PodTech

Guest: Paul Schopis – OARnet

Paul Lancour – PodTech

I am Paul Lancour with PodTech.net and right now I am here with Paul Schopis, he is the Associate Director of OARnet. Paul hello.

Paul Schopis – OARnet

Hello.

Paul Lancour – PodTech

Thanks very much for joining us, straight after that perhaps you could explain to our listeners what OARnet is?

Paul Schopis – OARnet

OARnet is the Research and Education Network for the state of Ohio, we are a project of the State Board of Regions and so we connect all of the Universities in the state to the Internet and the research and education backbones that are national.

Paul Lancour – PodTech

You are a part of the Ohio State University.

Paul Schopis – OARnet

Well, actually Ohio State of University acts as our fiduciary agent, we are actually a project of the State Board of Regions.

Paul Lancour – PodTech

So, as such, what does OARnet do for the education in the state of Ohio?

Paul Schopis – OARnet

Well, basically we provide them with their Internet connectivity and because we are an Internet2 GigaPoP we also provide services that’s — for example, most of the commodity providers toady do not carry such as IPv6 and Multicast and we also support Jumbo Frames.

Paul Lancour – PodTech

That is for not just the higher education in the state of Ohio, but that is for all of education in the state of Ohio.

Paul Schopis – OARnet

That part of it is just for higher education, OARnet also run the things called the TFN, the Third Frontier Network, which is a facilities based optical network. So, we actually run all of the optical services for ‘K-12′ as well, but we do not run their IP service.

Paul Lancour – PodTech

So, is there a way that you can give us an idea of what the breadth of clients is that you serve, how many people are involved?

Paul Schopis – OARnet

Basically, as far as our IP services we support — right now we have 88 Colleges, and Universities online and we estimate our downstream users to about 2.5 million.

Paul Lancour – PodTech

So, that is a vast numbers of people with a vast array of needs form the network. I would imagine that would create some special problems for you when setting up a network.

Paul Schopis – OARnet

Well, it does to some degree I mean we have — just to give you an idea, our largest University in the state is the Ohio State University and their undergraduate population is about 55,000, right now. They are one of the largest Universities in the world and our smallest of college is the Rabbinical College down in Cincinnati it has 12 students. So, there is a quite a variety of needs and services that people need.

Paul Lancour – PodTech

So, what are some of the specific problems that you face when trying to build out a network like that?

Paul Schopis – OARnet

Well, basically our biggest problem toady has been and it probably kind of really ties into our routing architecture, is about in the states we have 10 schools that are regular members of Internet2 and those are usually in almost case a large state school with the exception of Case Western, which is a private college and so then I have another 27 or 28 schools that are, what’s called the SEGP participant in the Internet2 program. So, those schools have access to the resources that Internet2 provides, the rest of the schools do not have that access, so it really creates — it is probably one of the harder things we do here is that, we have to create routing environment where the schools that have access to Abilene and Internet2 get that access and those services and the schools that are not entitled to them do not get them.

Paul Lancour – PodTech

OARnet was founded in 1987, is that right?

Paul Schopis – OARnet

Correct.

Paul Lancour – PodTech

How long have you been with OARnet?

Paul Schopis – OARnet

I have been with OARnet ten years.

Paul Lancour – PodTech

So, that entire time, I imagine you have seen a lot of changes in not only the technology but what you have had to do in terms of building that network and the demands on it.

Paul Schopis – OARnet

Well, for example, when I first started with the organization I was a field engineer and actually about the first year or two I worked for it and I was putting in T1 lines for schools and back in those days, it kind of sounds funny to say that, but back in those days T1 was considered high capacity. We would get the high cap services center when we were having problems. So, today to give you an idea in our backbones, to run our own backbone at this point we are currently deploying 10-gigabit services, so it is millions and millions of times faster than what I originally worked with.

Paul Lancour – PodTech

What role has Juniper played in building out this network?

Paul Schopis – OARnet

Well, Juniper solved a really very specific problem that we had. Going back to what I was talking about having to support our traffic up and who is eligible for what service. That kind of scenario is commonly referred to in the routing world as the Fish Problem and what has Juniper allowed us to do is because they have an implementation of logical routers we are able to create using one physical chassis as a logical router that actually serviced our Internet2 clients. So, we could then have separate instances of BGP and separate routing cables and we basically paste all that together to get it to the right head end by using MPLS.

So, basically, what that allows us do is, right on the edge we do all this filtering and marking and the logical router also access the PE router for the MPLS function and we’ll go ahead and tag the packets, so that when it hits the core all the router has do in the core, it does not have to understand very many rules if it has got an MPLS packet or MPLS shim in the header, it goes ahead and uses the LFIB, if it does not then it just uses the standard IP routing cable and that is how we are able to make that all work.

Paul Lancour – PodTech

Why is this called the Fish Problem?

Paul Schopis – OARnet

Actually, historically and I was not at the meeting, but back when these kinds of scenarios were starting to evolve but particularly in the higher educations space for the regional networks like we are and someone was drawing the problem and it kind of vaguely, I suppose if you had enough beer it probably helped, but it is kind of vaguely resembled the Christians Fish Symbol and for some reason somebody said, allegedly, but he said, “Hey that looks like the fish,” and so…

Paul Lancour – PodTech

It stuck.

Paul Schopis – OARnet

It kind of stuck. So, it is commonly referred to as the Fish Problem.

Paul Lancour – PodTech

So, looking back over the ten years that you have been with OARnet, I imagine it would have difficult to predict where you would be today in terms of the ability of technology and what the demands on you would be and I am wondering if you could have anticipated this ten years down the road and maybe looking ten years ahead from here, where you think this kind of networking challenge will go?

Paul Schopis – OARnet

Well, I would say ten years ago, no I could not even begin to imagine what we were doing these days. Back in those days, as I said, it was all T1, it was all strictly IP. The data rates were much, much lower, things such as Multicast were a real struggle, it is pretty commonplace these days.

Looking out towards the future I tend to think that we probably will continue to see network virtualization, in other words the way we are using logical routers today and there are a lot of research projects that are taking place at the national and international level, that there is notion that we will have virtualization in both — not only — for example, like in servers today in data centers, but you will actually have routers that have provisionable sections that allow a user to on-the-fly grab a resource as well as on-the-fly very dynamic network building, so say for example, you have a multi-lambda network that has a certain number of the lambdas are in use all the time as they are today, but you have some set aside in reserve for special occasions.

So, say for example, if there is a special project that needs a very, very high data exchange between a couple of, say supercomputer clusters that you would on-the-fly crank-up using GMPLS or some morally equivalent mechanism. I think GMPLS is the leader right now in the game, but would basically crank up those resources and the real tipping point today for that kind of a technology happening really is just pricing. I think as components in the optical world tend to continue to come down, things like that will start to become physically practical. I think there are a lot of logical and research reasons one would not want to do that today, I think what prevents people from doing more of it, is the expense involved.

But, as optics and multi-flexors become commoditized that just like 10-gigabit Ethernet using an analogy here, just look at the way the prices have come down and because of that price shift 10-gigabit Ethernet is becoming very commonplace at this point. So, I think looking towards the future we will see that kind of a thing gong on.

Paul Lancour – PodTech

I am wondering if the reason that the prices have come down on some of those things and will continue in the future is because of the laws that are associated with such things with the speeds of delivery just doubling and doubling or is it also because people’s expectations are such that the market is driving this kind of innovation?

Paul Schopis – OARnet

It is a little bit of both, I think the expectation drives it and as soon as the business case gets to be made. I mean for example, 10-gigabit transfer even in the SONET world the OC192 became a reality when the price point to install OC192 started to become a better deal than deploying four OC48.

So, similarly this happened in the gigabit world of Ethernet and I think similarly — well for example, I mean the optics are coming down dramatically in price and a lot of it is simply supply and demand, there is enough of a demand to drive people to produce such mechanisms; for example, today one can even buy an ITU Grid of specific SPF for a router or a switch and that is basically a color reserved piece of frequency of light that — they are getting terribly inexpensive. I mean in the order of like $300-500, I mean just put in perspective, three years ago we built the network a part like that was selling for about four times that.

Paul Lancour – PodTech

And I imagine that as you say those prices will continue to come down and things will get faster and faster.

Paul Schopis – OARnet

Yeah.

Paul Lancour – PodTech

Thanks for taking the time out of your schedule to speak with us today Paul.

Paul Schopis – OARnet

Have a good day.

Paul Lancour – PodTech

Paul Schopis is the Associate Director of OARnet.

Copyright ©2006 PodTech.net. All rights reserved. Privacy policy

Share

Posted in: Connected Social Media, Corporate, Juniper Networks, Master of IT, Technology
Tags: ,