Talk about your delayed reaction. Carl Henning, who writes The Profiblog, has no doubt been perplexed by my failure (or more to the point, of either ODVA’s or Rockwell Automation’s) to respond to his March 15 post about relative market shares of Profinet and EtherNet/IP.
Actually, I was busy trying to validate the information. Carl pointed to a press release from Profibus International that in turn refers to a study from UK-based research firm IMS Research that shows the relative market shares within industrial Ethernet protocols as EtherNet/IP at 30 percent, Profinet at 28 percent, Modbus TCP at 22 percent, Ethernet Powerlink at 11 percent and EtherCat at 4 percent.
I had a conversation with ODVA who could not confirm the numbers in the report. Rockwell Automation has the report, but it does not have approval from IMS to release information. IMS pointed me to a report of its press release in the online Industrial Ethernet Book. Then I found its press release here although it contains very little detail.
So, here goes a bit of an analysis. First, standard Ethernet TCP/IP has about 60 percent of the manufacturing market. Some might think that this is because of corporate connectivity, but several companies–notably National Instruments and Opto 22–have been proponents of standard Ethernet for years. So the specifically industrial protocols are dividing the remaining 40 percent.
Next, IMS said that Profinet was growing at 8.7 percent compound annual growth rate (CAGR) while EtherNet/IP was at 7.1 percent CAGR, and Modbus TCP was very slightly declining. By projecting these figures, Profinet looks to overtake EtherNet/IP within five years as the market share leader. If I were in the Profinet camp, I’d tout those numbers, too. But if I were an investor or outside observer, I’d take all projections with a large piece of the proverbial grain of salt.
New shipments of industrial Ethernet nodes declined in 2009 relative to 2008, but IMS expects 2010 numbers to grow back to 2008 numbers with 10 percent CAGR beginning in 2011.
This is the first that I’ve seen any kind of independent numbers (although I really don’t know the methodology of IMS–it could have just taken the word of the various industry organizations), and I’m frankly surprised that Profinet is that close to EtherNet/IP. Not because of any technical or marketing deficiency. More because of Profibus market penetration vis-a-vis DeviceNet.
The real takeaway? Just as I’ve been preaching for more than seven years–Ethernet is the industrial network of the future. And the future is just about here.
My thoughts:
1. As long as my suppliers are financially viable, I don’t care about market share, only a good match for my requirements.
2. I suspect it will be at least 5 (probably >10) years before Ethernet dominates over non-Ethernet buses
3. Just like standard Ethernet beats ALL industrial Ethernet protocols combined, I’m pretty sure non-fieldbus I/O and motion is a pretty big category, almost certainly bigger than any single fieldbus
4. Standard Ethernet isn’t really standard, because you have to figure out how to talk to each piece of equipment (e.g. each smart camera type, each robot type, each temperature controller type, each servo drive type, etc).
5. How are they counting Ethernet nodes? Equipment that is actually being used with a industrial Ethernet protocol or equipment that can use one (but might not be) — if the latter, I suspect they’re missing a lot of equipment. For example, I’ve got a Galil DMC21x3 that can talk Modbus/TCP (but doesn’t) and some Cognex smart cameras that can talk Modbus/TCP, Ethernet/IP or IIRC Profinet, but are currently communicating via telnet and ftp. Should these devices count towards Modbus/TCP, Ethernet/IP, and/or Profinet totals?
6. According to those stats, it looks like Powerlink is the clear real time Ethernet leader (unless Profinet IRT is >40% of Profinet nodes). I’m also dubious about the Powerlink vs EtherCAT numbers; EtherCAT seems to have signed up at least as many big motion players as Powerlinkl; I’d expect the numbers to be a lot closer.
7. Marketshare is also really going to vary with the application. There’s no way Modbus/TCP, Profinet, or Ethernet/IP is anywhere close to Powerlink or EtherCAT in servo drives. In the process space, however, I’m pretty sure you won’t see much Powerlink or EtherCAT.
Whew! Thanks, Gary, my perplexity is now passed. Knowing that delayed reactions are undesirable in control systems and blogging, here’s my un-delayed analysis of your analysis:
Our PROFINET node count was 1,600,000 through 2008 as counted by an independent third party. We count only automation devices, not Ethernet switches or other infrastructure devices. We don’t even count Ethernet switches that have PROFINET IO capability built in (http://www.us.profibus.com/wordpress/?p=185). At the SPS/IPC/Drives Show at the end of November 2008 an ODVA press release said they had 1,600,000 Ethernet/IP nodes. (I don’t know what they count as nodes.) Using the proportions from the IMS report: 1,600,000 nodes is 28% and Ethernet/IP is 30%, then the Ethernet/IP node count through 2008 was just over 1,700,000. I guess they had a really good December.
“[S]tandard Ethernet TCP/IP has about 60 percent of the manufacturing market.” This makes no sense to me. Ethernet plus TCP/IP is just the pipe not the pump. TCP/IP is not an application, so just standard Ethernet plus TCP/IP is not enough to move data from IO device to controller and back. Remember the seven layer ISO model collapses to four layers for Ethernet environments. The bottom two ISO layers are defined by IEEE as Ethernet. Layer 3 is IP (Internet Protocol) and Layer 4 is TCP (Transmission Control Protocol). Layers 5 and 6 are not used. But you still need layer 7, the application layer, to actually do something with the data that travels through Ethernet and TCP/IP. Layer 7 applications could be HTTP, Ethernet/IP, Modbus/TCP, and PROFINET. (“And” is correct because part of the beauty of Ethernet systems is that you can run multiple applications over the same wires.)
The question I don’t know the answer to then is this: If the marketing material says a product just uses standard Ethernet, what application is at the application layer?
Regarding the disproportion: You allude to the fact that PROFIBUS has maybe 4 or 5 times the installed base of DeviceNet while PROFINET is about equal to Ethernet/IP. The disproportion is really even worse when you consider that the Interbus folks with twice the installed base of DeviceNet decided to use PROFINET rather than create a new Industrial Ethernet. Two-part explanation – first, we started later. Second, there was more incentive for someone using DeviceNet (8 byte data packets at up to 0.5Mbaud) to move up to Ethernet/IP than for someone using PROFIBUS (244 byte data packets at up to 12Mbaud) to move up to PROFINET.
My conclusion is subtly different from yours:
The real takeaway? Just as I’ve been preaching for more than seven years – PROFINET is the industrial network of the future. And the future is here.
(Ok, maybe not very subtle.)
–Carl Henning
Thanks for the additions. I suppose "standard" Ethernet could use such application protocols as ftp, http or perhaps something custom. I have known of custom protocols for years that have many benefits if users limit themselves to a single brand.
Carl, I’m with you on the Ethernet Powerlink and EtherCat comparison. I was surprised at the disparity. Perhaps I’ll hear from Beckhoff on that one.
So, apparently EtherNet/IP and Profinet are in a close market share race. That will be interesting to watch for the primary companies involved and for component companies developing for the networks. I’m just gratified to see the Ethernet backbone thriving.
By the way, I’m told that Ethernet in process automation is complicated by a mix of proprietary application protocols and Foundation Fieldbus High Speed Ethernet protocol. Of course, Profinet has a play there due to the presence of Profibus PA. With Rockwell’s growing presence in process, will EtherNet/IP have a play there? No one has opened up on details. Maybe some process readers can chime in.
I can only offer a perspective from a very small geographic area — the Chattanooga, TN MSA. In this area the adoption rate of Ethernet/IP by existing Rockwell Automation customers is very high. Ethernet/IP has displaced virtually every other network on both new systems and upgrades to existing systems. Profinet is probably only holding it’s own due to a very large installation at the new VW facility that is currently under construction here.
Hey Doug, thanks for the input. Always appreciate hearing what’s really going on rather than what marketers tell me is going on 😉
Further on this subject, I heard from Graham Harris of Beckhoff (USA) and Joey Stubbs of the EtherCat association (USA) on the numbers. They responded by email and I’m not sure that they wanted exact quotes, but here is the gist of what they brought up.
They had no numbers to either confirm or dispute the relative market shares suggested by IMS Research. Harris did point out that Beckhoff has had standard Ethernet switches for years to which they have connected I/O devices.
They both pointed out some of the usual problems with all these market studies of networking–what is the definition of a node. Further, what is the source of the information. Carl Henning pointed to some sources for Profinet v EtherNet/IP. This may be as close as you can get. Harris wondered just what Rockwell Automation counted as an EtherNet/IP node.
These reports get out, so I feel like I should try to shed light on them. But market share data is extremely hard to corroborate.
By the way, we also get into debates on what Ethernet is. I believe that this is market research on industrial Ethernet application layer protocols that are carried by an Ethernet network. This also is not a technical debate on what can be carried on "standard" Ethernet, which I presume to be using common protocols such as ftp, http, snmp and the like. Whether those are capable of anything is beyond the scope of this discussion. (But would be a really interesting discussion on its own.)
The other thing that will be interesting to watch is what happens when Profinet IRT and CIP Motion (when products appear) start to gain momentum from their base of Profinet and EtherNet/IP relative to EtherCat and Ethernet Powerlink.
Those of you who are designing systems–what do you intend to use for motion? Keep the comments coming.