Goodbye Switches???

I remember about four years ago getting some severe criticism for an email I sent out that Ethernet would eventually replace DeviceNet, Profibus and the rest. His point was that we’ll never replace those things because they are linear topologies and Ethernet costs so much more because of the star topology.

We’ll guess what? I was pretty farsighted on this one. In case you don’t know it, there is a huge effort underway to do daisy-chained Ethernet. There are a couple of ways to do that. One is that embedding a switch right into the device. That’s a 3-port switch with one incoming port, one outgoing port and a port for the device. It’s not as fast as connecting 32 devices to a single switch. In that case, your message goes immediately out the port connected to the target device. Instead, in the linear Ethernet sytstem, you have to go through all the downstream or up stream devices to get to your target device. But in exchange for the more complicated, more expensive daisy chained node, you save a ton of wiring costs and the cost of the big switch. It’s a good tradeoff as your network is simpler.

Another way to approach this is to huge a very fast processor with dual Ethernet ports and embed the switch software right in the processor. That is a simpler architecture but the cost of the faster, more high end processor is offset by the loss of the standalone switch IC.  We’re doing a couple different applications in this area. Mostly related to conveyors and material handling. That’s the kind of stuff we do for our OEM, custom product development business. You can check it out here:There are, of course, Ethernet protocols with this stuff built right in; Profinet IRT, EtherCat, PowerLink to name a few.  I’ll have more to say about those in a future post.

JR