Grand Challenges in Networks for the Next 15 Years

jameshowison writes "Some of the researchers responsible for the Internet, including Bob Branden of ISI and David D Clark from MIT, have outlined what they see as the grand challenges for internetworking and computation in the next 10-15 years (PDF). The report from the IRTF's 'End-to-End Research Group' discussed the question, 'How might the computing and communications world be materially different in 10 to 15 years' and how do we get there? From a universal system for location, to small-area networks, to operation in time of crisis, software radio and an agenda to reduce the energy required for communications this document tries to imagine what will be like packet-switching was for the past 15 years."

Comments

The Devil is in the Details

[suso]

Appearently, using HTML for documents is still a major challenge.

It only takes one person or company to implement things wrong, break protocol and then you have a mess. That is the grand challenge.

Re:The Devil is in the Details

[sp3tt]

"using HTML for documents is still a major challenge" Yeah, Microsofts helps making that a problem with IE.

Are you sure

[elid]

that this isn't one of those randomly generated MIT papers?

Re:Are you sure

[spac3manspiff]

haha, 200 years of credibility just got flushed down the drain.

My challenge

[Anonymous Coward]

this document tries to imagine what will be like packet-switching was for the past 15 years.

I'm trying to imagine what this sentence means.. and it might take me 10-15 years.

Ubiquitous, seamless, personalized spam

[Animats]

The major Internet applications, by volume, are spam, piracy, and advertising. This trend will continue. By 2020, 98% of all Internet traffic will be illegal in some way.

Already there

[samael]

They claim there isn't an emergency broadcast system - but we have Slashdot! The second anything big goes wrong, there it is!

Are we talking about the US of A?

[Beatbyte]

Because it's still going to be the WAN from LAN network that we'll be working on forever.

I've got a LAN setup running 200x as fast as the fastest WAN/Internet connection readily available (minus a special order and uber expensive DS3). And at the pace we're going, the US is getting slower and slower as far as the Internet connections go.

Right now I can completely rewire my office and home for $5k with state of the art, high end network components and have it done in less than a week. I can't get close to those speeds with my net connection for 4x that price ($20k/year).

That being said, there is still hope somewhere

Re:Are we talking about the US of A?

[Wesley Felter]

The document focuses on technical challenges, not business or political ones. But you're right that technical innovation is useless unless there is a business and political climate that can foster it.

Re:Are we talking about the US of A?

[Dun Malg]

Right now I can completely rewire my office and home for $5k with state of the art, high end network components and have it done in less than a week. I can't get close to those speeds with my net connection for 4x that price ($20k/year).

You say that like there's no difference between a LAN and a WAN. The reason your LAN setup is so much cheaper is that none of your cable runs are THREE MILES LONG. You think teh intarweb runs over 100BTX Ethernet cable everywhere?

Re:Are we talking about the US of A?

[tomstdenis]

Not only the distance but latency. Switching 1500 byte packets locally between two computers is trivial.

Try that with 300,000 subscribers ...

Tom

Re:Are we talking about the US of A?

[Beatbyte]

That's my point. Your excuse is distance. My point is, I don't care. I want speed. We need to quit focusing so much effort on making LAN's faster and focus on WAN/Internet connections.

As far "teh intarweb" you speak of... nope, I don't think it runs on "100BTX Ethernet cable".. I've been in the ISP business for 10 years now and I'm pretty familiar with both ends of the Internet. The first being the provider end. The second, being the customer's end. Considering the customers pay my bills, I'm more worried about providing them with what they want.

IPv6

[Anonymous Coward]

The biggest challenge will be moving the entire internet onto IPv6

Wait a second...

[bungley]

In 10 years, our communications infrastructure should be based on an architecture that provides a coherent framework for security, robust operation in the face of attack, and a trustworthy environment or services and applications.

Wasn't that what it was designed for in the first place?

Advanced interconnectivity for inter-personal comm

[mekkab]

We need to improve interpersonal communication via computer internetworking. And until Punch You In The Face over Ethernet (PYITFoE) is widely available, we will only ever scratch the surface of the rich tapestry of human interaction.

Location technology

[jessecurry]

The little section about location technology was very interesting. I love using my GPS, it has opened up a new sport to me and allows me to do some very interesting things, but I am bothered by the fact that it only works outdoors. Having a GPS-like systems that worked everywhere will be very cool, and integration with existing devices might bring the "smart home" I've been hearing about for the past 15 years into reality.
The one part missing from my home automation system is the ability to autonomously process input. I have to use a remote control for events that aren't based on a repeating schedule. It would be nice to be able to walk into a room and have my wrist watch alert my automation server as to my whereabouts, then have the lighting dynamically adjust to me.

Re:Location technology

[Detritus]

You might have better luck with ranging based upon propagation delay. If you know the locations of the base stations, and the transmitted signal contains a time reference, you can measure the delay between the transmitter and receiver. This assumes that each node has an accurate clock. Another approach is for the base station to transmit a carrier modulated by a PN sequence. The mobile station takes the output of its receiver (the PN sequence) and feeds it to the modulator in its transmitter. The base station compares the transmitted PN sequence to the received PN sequence to make a delay measurement. After subtracting the known delays in the system, this gives you a round-trip delay measurement.

What we need to do...

[jgold03]

Switch to IPv6

Multimedia "over IP" will not become mainstream without virtual circuit technologies. Also, we are being lazy and letting NAT take care of the lack of addressing provided by IPv4.

Re:What we need to do...

[jgold03]

I would think that rather than virtual circuits what we need is effective flow control (but maybe that is what you mean....). Flow control is pretty much the same thing as virtual circuits. When two end hosts need a path with a certain amount of bandwidth across the Internet, the routers have to maintain state information about that connection in order to provide it's guarantees. IPv6's proposed solution is to add a flow # to the packet header.

WQos, .mil, trusted computing, etc.

[G4from128k]

A few ideas:

1. QoS: The wider use of quality-of-service metrics to regulate bandwidth/latency/drop-rate will spread from backbone to the backplane. QoS will be assigned not just to packets or network streams, but extended to applications, processes, and threads.
   
2. authentication-intensive network: Anti-spam, anti-phish, anti-piracy initiatives will deanonymize the network. Expanding liability may force commercial providers of network infrastructure to adopt so-called trusted computing initiatives. Counterfeiting a header may become crime similar to counterfeiting money because both crimes degrade the public trust in the system.
   
3. militarization of networks: When will network security become so important to the national interest that the government deploys .mil computers to DDoS offending servers. If the economy runs on the net, someone will become the defender of that infrastructure.
   
4. physical layer/application layer dichotomy: Currently the value of the network is in the application, but the cost is in the physical layer. This lead to the problem of price wars among infrastructure service provider or the war over municipal wifi. Perhaps an alternate approach would more closely link the value and costs of networking.
   
5. Multiple IPs per device: I wonder if the move to multiple cores will push systems toward multiple IPv6 addys per machine. Technologies such as IBM's cell architecture support the potential for multiple OSes running on a single hardware platform. With such a large IPv6 address space, it may be easier to give each running OS instance its own IP address, rather than try to share an address and try to use a meta OS to share network resources. This, in turn, may lead to a proliferation of addresses that fill the larger space.

Solutions without Problems

[ItWasThem]

The main issue I take with this paper is that it proposes a series of solutions without talking about any relevant application or problem that it will solve except for in an occasionally very generic way "We need better security" for example.

That and the fact that it seems to have been written with the longest most convoluted sentences possible.

Major change happens when an intelligent person solves a very real problem in a way that seems obvious once it's completed but that few others would have come up with.

This paper starts by dissing incremental improvements and then goes on to rehash... wait for it... incremental improvements. How can you compare "better security" to Packet Switching in terms of revolutionary technology?

In my opinion major advances in the next 10-15 years will be driven by content-based applications. Technology is cheap and is becomming a commodity. It will not make any more major leaps until there is a content driver and industry to take it there.

For example, when we can all print flat panels for wall paper what will we have to display on them? An entirely new content and distribution industry will emerge to fill these and other voids and THEN technology will again stride ahead.

Just my .02

Re:IPv6 isn't a big deal

[billstewart]

IPv6 is useful, and at some point we'll need the address space, but basically until Cisco and Juniper make routers that perform well using IPv6, nobody feels motivated to move wholesale - almost half the IPv4 space is still unused.. Microsoft is doing a bunch of IPv6 work that'll help the chicken&egg problem in a couple of years, but without a killer app, there's no real motivation.

The big problem I've seen with IPv6 is that its goals not only included bigger address space, which we've been able to slack off by using RFC1918 private space, firewalls, and NAT, but it also promised to do Really Cool Things to make routing infrastructures more scalable and better behaved, and that doesn't appear to have panned out yet. That means that not only do routing tables get bigger because the addresses are longer (which you fix by waiting for a couple of year's of Moore's Law to fix memory pricing), but there's likely to be a repeat of the "IPv4 Class C Address Swamp" which nobody wants, or the "Upstream-Provider Non-Portable Address Space Lock-In" features which customers don't like, and which makes multi-homing for reliability much harder. And that doesn't seem to have been done yet.