Startup Promotes Switchless Networks

A small UK startup is promoting a network architecture that could upset a lot of big equipment vendors if it's successful -- a bidirectional "bus" network that requires neither switches nor routers.

Called Pipistrelle Ltd. (Website still under construction), the company is still in its early stages. It's waiting for a technology license from Lockheed Martin Corp., which covers the basic architecture. Lockheed also intends to license the technology to a newly minted startup in Florida (name unknown) to cover the American market.

Lockheed developed the technology some time ago but didn't want to commercialize it for telecom networks, according to Pipistrelle's managing director Dick Winchester. And rather than give the idea to a company already in the communications business, it decided to encourage startups that would be solely focused on what it calls "fiber-optic bus wavelength division multiplexing" (FOBWDM).

The main feature of FOBWDM is that it's all optical, meaning that there are no electronics in the path of the data. That allows the core infrastructure of the network to be protocol- and data-rate-independent. Winchester claims that FOBWDM is suitable for speeds up to and including 40 Gbit/s.

Being a bus system, it means that the data does not need to be routed to a particular destination. It is broadcast through the system, and when it passes through the destination node, it gets recognized by the equipment and picked up. As a result, FOBWDM doesn't require switches or routers.

Instead, what's called an "optical bus interface module" (OBIM) is placed at each node in the network. This contains passive splitters that tap signals traveling on the network and place new signals into circulation.

An OBIM feeds signals into an interface card, which contains some kind of magic to ensure that all the signals on the network are kept up to strength optically. "You would think that the signal-to-noise ratio would worsen, but in fact it doesn't," notes Winchester. This observation was made on a demonstrator network built at Lockheed's plant in San Diego, he adds.

On a small scale, the FOBWDM idea has already proven itself. Lockheed has put the idea into practice as the communications nervous system inside two NP-3 testbed aircraft. However, putting communications systems comprising roughly 100 feet of cable into military aircraft is an entirely different thing from building a national or regional network covering hundreds of miles. Winchester claims that it would be possible to build a FOBWDM that covered the entire U.K. but can furnish little evidence to prove that it would work as advertised.

Pipistrelle also faces the not insignificant hurdle of convincing people to buy into a rather radical idea that would require drastic changes in their network infrastructures.

By its nature, the bus network is bidirectional -- signals travel in both directions along the fiber. That immediately rules out the possibility of using Erbium Doped-Fiber Amplifiers (EDFAs) in the network, since they contain isolators that only permit signals to pass in one direction. Therefore, most existing fiber plant is unsuitable for FOBWDM.

Other startups that have challenged mainstream technology, such as, for example, CodeStream Technologies with its OCDMA (optical code division multiple access) modulation scheme, have met with resistance, even though there may have been certain advantages to the technology (see CodeStream Goes Under).

Right now, Pipistrelle comprises only three people: Winchester and two others, whom he is not at liberty to name, since they haven't left their present employers. The company is registered at Insch in Aberdeenshire, Scotland, "where the air is clearer and houses are considerably cheaper." It remains to be seen if this batty idea will fly. The first task, apart from waiting for the license agreement, is to identify where the opportunities are for FOBWDM networks.

— Pauline Rigby, Senior Editor, Light Reading
Page 1 / 3   >   >>
brian.murray 12/4/2012 | 11:06:36 PM
re: Startup Promotes Switchless Networks
out of curiousity metroman ... where in the UK do you see as being the hive(s) of the the fibre related industry?
Pseudopersonality 12/4/2012 | 7:32:06 PM
re: Startup Promotes Switchless Networks This idea is screaming out for a startup to figure out!

Looks like a local area network possibility like an...aircraft.

Have fun
santosh1914 12/4/2012 | 7:32:05 PM
re: Startup Promotes Switchless Networks The idea of a fiber bus seems interesting. Its lot like Ethernet with few differences, those due to the availablility of multiple independent wavelengths. Here is one way this network may work :
The data-transparency in optical network is due to the fact that a switch can connect an input port to an outport by looking at the wavelength only. The cross-connect state for a particular state is decided in advance through signaling. In FOBWDM also one channel can be devoted to exchanging signaling messages. This channel may be used to decide which pair of transmitter and receiver or a number of receivers (in case of multicast) will use a particular wavelength. The transmitter will have to check for all the free wavelengths on which it is capable of transmitting and then reserve a wavelength that the receiver(s) can receive on. The transmitter and receiver(s) will have agree on a common framing type.
Please note that the above scheme does not involve a CSMA/CD like mechanism except on the signaling channel. Its possible to use a scheme like CSMA/CD on data channels also but will be more complex.
Any more ideas or comments on above?
Its more improtant to find out its benefits w.r.t existing network topologies than figuring out how this may work. Here are my views:
1. This will use less fiber.
2. There is no need for a switching device. We still need mux/demux and amplifiers. So, there is sigificant cost savings given the high cost of optical switches.
3. The optical signals will incur less loss since they do not have to pass through a switching device.

1. The scalability is a big issue because the bandwidth is being shared among all nodes on the bus. As the number of nodes increase on the bus the number of connections that can be established per node will decrease. Increasing fibers on the bus will cause the number of transceivers to increase at the nodes. So, we may seem its application in networks of limited size (of the order of tens). The need of switches will still be there.
2. Reliability is a big issue unless we use SONET like protection mechanism of devoting half the bandwidth for protection.

DickW 12/4/2012 | 7:32:05 PM
re: Startup Promotes Switchless Networks Hi

Giles first... FOBWDM does not suffer from Ethernet style collisions...

Santosh ..... Your "advantages" points are pretty much spot on but there are others..

Your "disadvantages" are way off beam though.. The scaleability of the technology is much larger than you suggest.. certainly up to about 1200 "nodes" and perhaps greater.. AND, I can assure you only need a switch if you have to connect to the outside world... As far as reliability is concerned there is no need for the sort of protection you suggest..

[email protected]
giles0 12/4/2012 | 7:32:05 PM
re: Startup Promotes Switchless Networks Isn't this a subset of what George Gilder was talking about 9 years ago in "The Coming of the Fibersphere"?

"You can send all messages everywhere in the network, include all needed codes and instructions for correcting, decrypting, and reading them, and allow each terminal to tune into its own messages on its own wavelength, just like a two-way radio. When the terminals are smart enough and the bandwidth great enough, your all optical network can be as dumb as a stone."

a nice idea - but maybe a wee bit utopian? ;-)

From the name of the technology (which has WDM in it) I get the impression that Pipistrelle would assign each device a dedicated wavelength to receive on or to transmit on? The former would presumably require tunable lasers in the transmitting nodes and would potentially suffer from Ethernet-style collisions, whilst the latter would require a node to listen to all frequencies simultaneously?

anyone got any better idea as to how this might work?

giles0 12/4/2012 | 7:32:02 PM
re: Startup Promotes Switchless Networks "Giles first... FOBWDM does not suffer from Ethernet style collisions..."

so can I assume from this that you don't dedicate a wavelength to each receiver?

do you then dedicate one to each transmitter - or is it something more complex?

If the former then I can't see how you'd scale to 1200 nodes?
veemee 12/4/2012 | 7:32:01 PM
re: Startup Promotes Switchless Networks Also there are going to be a limitations. The server will be a bottleneck. (Servers that allocate bandwidth to a particualr session, if things are not going to be CSMA/CD)
yikes_stripes 12/4/2012 | 7:31:59 PM
re: Startup Promotes Switchless Networks The MIL STD 1773 bus was being looked at for possible expansion into the WDM realm at a place I worked at a while ago. There is alot of COTS available for this stuff. Just rip out the LED and put in an ITU laser and away you go. Lot's of fun in your garage.
santosh1914 12/4/2012 | 7:31:58 PM
re: Startup Promotes Switchless Networks Hi Dick,
Can you throw some light on what is the difference between a SONET ring opened up to look like a bus and the FOBWDM? Why should someone deploying a SONET Ring today consider FOBWDM as an alternative? I see the following points:

Let us call the SONET as switched approach and FOBWDM as switchless approach to building WDM networks on a bus topology.
1. Both approach will use same amount of fiber.
2. The switchless version (FOBWDM) will cost less.
3. The switchless version will need less number of amplifiers for the same distance or receivers with less sensitivity and/or transmitters with less power.
4. Now let us compare their scalability: Let us assume there is one pair of fiber in the bus with 40 lambdas. Let us devote 1 to control channel. The maximum number of connections that can be established in the switched version of the network will be (n-1)*39 where n is the number of switches. In the switchless version the maximum number of connections that can be established in the entire network will be 39 irrespective of the number of nodes in the network and these 39 connections will be shared by all the nodes. This is because increasing the broadcast domain reduces the wavelength (or more generally channel) reuse. So, as we increase the number of nodes in both versions of the network we can see that the number of connections that can be established per node reduces drastically ( (n-1)*39/n in the switched network vs 39/n in the switchless version). This is what I was referring to as scalability issues.

Santosh Kumar
Ohio State University
PhilMorrison99 12/4/2012 | 7:31:58 PM
re: Startup Promotes Switchless Networks That reminds me of a similar story about another couple who used to have "lot's of fun" in their garage. The eventually formed a company called cisco or should I say Cisco :-).
Page 1 / 3   >   >>
Sign In