Design Con 2015
Breaking News
Blog

Debunking MOST vs. Ethernet Debate

Choosing the right transport for automotive data
NO RATINGS
Page 1 / 6 Next >
View Comments: Newest First | Oldest First | Threaded View
Henry Muyshondt
User Rank
Rookie
Re: This debate has occurred before
Henry Muyshondt   10/3/2014 5:40:36 AM
NO RATINGS
My point was that you don't really need to choose one technology or the other. With MOST, you can choose to use packet switching and IP communication for the data streams that benefit from it, while you can also circuit switching for others, like low latency/high determinism driver assistance systems. The physical layer is specifically designed to handle both synchronous and asynchronous traffic.

There are already over 150 million MOST nodes on the road, in over 172 different car models. The adoption of IP communications doesn't mean having to throw out the existing infrastructure, since it can already move Ethernet frames unmodified. In fact, the latest generation of MOST has a dedicated Ethernet channel, just for IP communications.

Another aspect of this discussion is what the definition of Ethernet is. Some use it to mean a specific frame format, others the IEEE 802.3 physical medium (which is where the term actually originated). I believe the majority of people really mean IP communications, as you describe them, and that has nothing to with the actual physical medium used. IP communications flow over many different media as they go from one system to another. I also believe the focus of car communications, especially in the infotainment domain, is moving towards communication outside the vehicle, as more data moves to and from "the cloud". This will most likely involve IP communication over a wireless link (which is not Ethernet in the sense of the 802.3 physical layer). There is definitely a need for IP communications though there are lower overhead mechanisms for other applications.

Finally, as you mention, Ethernet has always relied on oveprovisioning to overcome its challenges. This can be a problem in the harsh automotive environment, as faster data rates mean higher electromagnetic interference and susceptibility. Current car architectures seek to "right-size" the physical media used. They don't want to put in a Gb pipe if all they need is to move 100Mbit/s with low latency and full determinism. They also don't want to put in a 10Gb pipe if what they want is a couple of 1.3 Gbit/s or more uncompressed HD video streams. Instead, they need a pipe that can best handle the different data types simultaneously.

krisi
User Rank
CEO
Re: This debate has occurred before
krisi   10/2/2014 7:06:27 PM
NO RATINGS
That is largely true Bert...still, the only reliable way to call 911 is to thru stationary phone...number of 99999 is still larger by at least 1 compared to wireless networks...so I am not taking any chances and keeping my wired phone...would I want my car use Ethernet? Not really but likely I will have no choice in that matter

Bert22306
User Rank
CEO
This debate has occurred before
Bert22306   10/2/2014 6:34:18 PM
NO RATINGS
Back in the 1990s, the telephone carriers were also debating these questions. Should the telephone standards stick to their historically isochronous media? Or can they make effective use of packet-switched IP and Ethernet backbones?

Packet switching won out, in general, and is taking on a bigger and bigger piece of the pie as time passes. I use voice over IP/Ethernet for my telephone service at work, I  watch TV over Internet Protocols and Ethernet (or WiFi) at home. Any glitch I do occasionally notice is always caused by interruptions in the ISP or enterprise connection, as in total lapses of service when there's a hardware fault in the system, not by timing issues between packets caused by mere traffic congestion. Such hardware failures can occur over any medium. So I think some of the concerns here are overstated.

We have run live video streams over IP/Ethernet control system networks, initially in demos and then in the field, with absolutely no problem. In fact, we have even run up the traffic volume to 100 percent of capacity, to demonstrate that nothing falls apart. The worst case delays become slightly longer, up in the couple of millisecond region as opposed to a millisecond or less end to end, but nothing catastrophic happens. Doing the math shows why, but people have to see it to believe it.

There are many control systems now, in factories, ships, and other platforms, that have migrated to packet switching and IP over Ethernet. In fact, some of the traditional industrial bus protocols have been layered directly on top of IP/Ethernet, such as Modbus over TCP or BACnet over UDP. Or alternatively, gateways can be used to transfer IP traffic to and from these legacy buses.

So overall, my prediction would be that car control system communications will follow this same evolution. All of the low data rate traffic, such as that on CAN, would most likely operate without a hitch on the much, much faster Ethernet nets. Some Ethernet nets will likely be segregated from others, e.g. critical controls separated from infotainment. And then over-provisioning the Ethernet networks ensures very short delays and virtually synchronized operation. Honestly, you won't notice any of these supposed congestion issues. And such over-provisioning will be far easier and cheaper to do, than it would be to upgrade the older timed network schemes for the higher data rates required in the future.

It takes time to change the culture, but all you have to do is look at other examples where this has happened. Starting with the telcos.

Flash Poll
Like Us on Facebook

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
EE Times on Twitter
EE Times Twitter Feed
Top Comments of the Week