Should routers and modems talk to each other ?

There is a problem looming in the marketplace – more and more people are moving to wireless communication solutions, and Internet Protocol (IP) is pervasive. This is all fine today, but the problem arises when we need to get more from less, as wireless spectrum disappears. In the satellite world solutions such as DVB-S2/ACM vary signal power to meet the current channel conditions (due to clouds for example). On the ground many wireless modems tune to meet the current channel conditions as mobile devices move around dense cities. Again this is fine – however in nearly all networks there is a router, and in order to ensure some level of Quality of Service (QOS) the modem has some level of traffic shaping enabled to prioritise delay or jitter sensitive traffic such as VoIP. This invariably means that the router traffic shapes the traffic going to the modem to a pre defined rate. So when the modem changes speed the router is blissfully unaware and continue to spit out traffic at the old rate regardless of whether the modem has increased or reduced its offered rate. The ubiquitous use of Ethernet has caused a disassociation of not only modem link speed, but also link state and a few other parameters.

So whats the answer – well its actually quite simple – routers and modems should talk to each other. The actual mechanism doesn’t really matter that much, but what is important is that they do talk. This is often called cross-layer design, where a modem which traditionally operates in OSI layer 2 starts to talk out of band to the router that is shaping in OSI layer 3 data rates.

This problem isn’t confined to wireless and satellite modems, modems such as ADSL that auto tune will also need the same feature in the future. It also applies to tunnels such as VPN tunnels where the network state is unknown to the “red” side of the tunnel.

In my previous life many moons ago, a colleague and I tried to start fix this problem. We wrote a few IETF drafts with limited sucess.

http://tools.ietf.org/html/draft-floreani-ancp-wireless-00

Click to access draft-wood-dna-link-properties-advertisement-01.pdf

However there is a real problem – there exists no standards body that this problem neatly fits into. The layer 3 at the IETF don’t really do much layer 2 anymore. The ANCP people in the IETF only look after DSL (and maybe some wireless networks in the future) but they don’t consider the CPE router as their remit. How can we form a standard interface in which satellite modem vendors, WiFi and WiMax vendors, Military Radio vendors, DSL modem vendors all work together with the multiple Layer 3 vendors when there is no standards body that includes these players at present ?

the technical solution is easy – the politics is a nightmare.

 

These views are my own and do not represent the views of my employer

Future of Modem Vendors.

I have this idea, and the idea isnt pretty if you make your money out of selling modems. If I am bonkers please let me know.

So IP seems to be the convergence layer of choice – no debate. Nearly every modem these days has migrated from just providing a serial link to having an ethernet interface and being able to interpret IP traffic natively. Modems are making descisions based on observing the channel as before, but also by analysing the IP traffic and optimising resource allocation accordingly. Smart modems are carrying out some level of crosslayer activity, taking layer 3 information to make layer 2 and 1 descisions. If we take this a step futher modems are necessarily becoming more and more layer 3 aware.

This is where the problem comes as these modem vendors quickly hit a fork in the road – do they continue to re invent the wheel (ie layer 3 routing, QoS, IPSEC, call control and much much more) and bump heads with the big layer 3 vendors such as Cisco, Juniper and Huawei or do they succumb and considering partnering with one or more layer 3 vendors.

This is where the crunch begins.

1. At a minimum the modem vendor will need to start to implement the router modem interface that I have spoken about in a seperate post. The interface is necessary to allow smooth operation of an IP network in a dynamic environment. But what happens in this world to the modem vendor ? For the interface to be of use the modem relinquishes its layer 3 capabilities to the router, leaving the modem to do what it does best, deal with the channel. Modem vendors will then have one of their competitive edges removed and they can only resort to cost, QoS capabilities and channel efficiency as their selling points. All modem vendors will soon be very competitive. You could even say modems become commoditised.
2. A modem vendor could take the path of building modules to go in a router vendors chassis. Whilst this seems feasible at first what inevitably happens is that the roadmaps of the two vendors diverge or get out of sync causing issues. Support of hardware provided by two vendors is always a nightmare. The router vendor also then gets multiple requests from the vast array of bespoke modem vendors to integrate their hardware, leaving them with a much larger engineering and support problem. The router vendor ends up annoying more modem vendors that it pleases.
3. So the router vendor inevitably starts dreaming of a better world where they produce some common hardware and the modem vendors port their code to this hardware. Technology has reached the stage where a couple of good DSPs and FPGAs can do a very good job. The specific RF front ends could in many cases be daughter cards to the generic hardware. This is no different to the world of software definable radios. Now the modem vendors are desitined to be software vendors losing their ability to compete on hardware differentiators
4. One more thing – whilst the layer 3 world has been living in the world of standards for some time, in many cases the modem world is new to this. WiFi is an example of modems in a standards world, but in the satcom area there are many many standards out there. Satcom modems are moving to DVB, but even then they only use this for traffic going to the CPE modem, return path traffic is often still proprietary. This anachronism wont last forever (dvb-rcs is a good focus point here). Once each medium has one, or a selection of air interface standards, then the modem vendors are now left without proprietary hardware and interface implementations, and can only differentiate based on their code performance.

If you think that is bad for modem vendors, I suspect there is another convergence in the wings – between broadcast services, point to point, and broadband services. But thats a subject of another post.

While the story sounds compelling to me – its not something that I see happening veryquickly, mainly as the market is so diverse it is not attractive to big players to go after in one effort.

These views are my own and do not represent the views of my employer

Ciscos second router in orbit launched today

The Cisco IRIS payload sucessfully launched today (23/Nov/2009) into GEO orbit on an atlas V rocket. The first router, a Cisco 3200 onboard the UK-DMC LEO satellite (launched in Sept 2003) is still working 6+ years after launch. This new router will allow further testing of the applicability of IP in Space. Whilst it is still very early days, and the IRIS payload is still a demonstration payload, it will allow customers and governments to test radical new ways to deploy IP services in space.

Interestingly I was told today its 4 years to the day that the concept of IRIS was first discussed with the eventual sponsors of the IRIS JCTD. Maybe the delayed launch was always meant to be. This emphasises the possible quick turn around in using COTS widely available equipment and software in the production of space hardware and services.

 

These views are my own and do not represent the views of my employer

Hello world!

Welcome to WordPress.com. This is your first post. Edit or delete it and start blogging!