ATM Node in a cog

Todd Marshall

I'm beginning to look at the feasibility of putting an ATM node in a COG. Then to put it into 2 COGS and have them communicate. Then into 7 COGS to form a mesh network. Then to multiple Prop 1s to expand the network. Then to add wireless (ZBEE? or more desirably UWB) to give flexibility to the network. Then to add GPS (or static location) information to the nodes to give more network flexibility. Then to add a WIZNET to give Layer 3. Then maybe to install my GLEE programming language (see WithGLEE.com) in a Cyclone V LPGA to give higher layer functionality.

Does anyone see a reason an ATM node cannot be put on a COG ... or even on a Prop1 or anticipated Prop 2? ATM brings up nothing when I do a search of the Parallax site.

If anyone else has interest in such an effort I welcome collaboration.

kwinn

I see well over a dozen meanings for ATM when I google it. COULD YOU BE A BIT MORE SPECIFIC?

abecedarian

I would think that unless you have some compelling reason to implement ATM, don't do it. IP routing is where the tech is going, particularly now that IPV6 is in play.

jazzed

Asynchronous Transfer Mode ?

Todd Marshall

ATM: Asynchronous Transfer Mode

https://en.wikipedia.org/wiki/Asynchronous_Transfer_Mode

Todd Marshall

I know that. And it's even progressed to the ridiculous point that they're inventing ways to run ATM over IP-MPLS. This clearly shows the eye has been taken off of the ball. IP (whether IPV4 and IPV6) is a layer 3 protocol. In the beginning IP ran almost exclusively over ATM. Then Cisco got involved politically.

ATM is a layer 2 (and 2.5 ... with the AAL versions) protocol. I was in the business when we lost our way.

Regardless, does anyone see a deficiency in a COG that precludes implementing an ATM node in one?

jazzed

I worked on ATM (we called it Automated Teller Machine in our specs) for several years, then moved on to SONET/SDH.

Many times I've considered Propeller for doing TDM (Time Domain Multiplexing) because it can do TDM ... except that the darn thing is too slow to produce an OC1 rate. This is one reason I asked for SERDES in the new chip. Without SERDES, even the new chip would not be able to do OC1.

So, while you can do some ATM cell communication and framing etc... with Propeller, where would it be applied given the speed restrictions?

xanadu

ATM and token ring are two really cool networking technologies. My first Cisco cert was quite a bit of ATM/Ethernet. I still have some ATM modules laying around. They seem to have a lot of chips on them, makes me think it would be possible to do on a P1 with external hardware.

Todd Marshall

I think in a widely dispersed mesh network (where there is no backbone), OC1 speeds are not required. IP can only make about 20 hops in the 1/8 second required to make realtime voice transmission tolerable. Thus, it needs a backbone to make the long haul and that means concentration and trunking.

ATM can hop at three or four orders of magnitude faster than IP and is naturally connection oriented (yielding security and efficiency benefits). The challenge is to keep thousands of nodes alive (or re-routable) in a "connection" for the length of the transmission. That's more difficult than keeping 20 alive ... but still possible. Heck, IP does it on a per packet basis.

Todd Marshall

When I was exploring this about 10 years ago there were single chip solutions on the market for ATM nodes. They don't exist now because ATM has been driven out of existence ... I think wrongly.