FUD Dissection Part 1

Posted by & filed under Uncategorized.

You may recall that I recently received a document comparing Ethernet/IP and PROFINET that was passed out at a Rockwell Automation Ethernet/IP “Lunch and Learn” according to the person who passed it on to me. (I initially blogged about the piece here, prompting a response from Gary Mintchell which prompted a few comments to Gary’s observations.)

It’s time to correct the mistakes in this three-page Rockwell Automation document.The document compares Ethernet/IP and PROFINET, but not accurately.You can find page 1 of the document here (you might want to print that to guide you through the details which follow).

Scalpel in hand; let the dissection begin – the first section is titled “Established” and has some statements about Ethernet/IP (EIP):

  • EIP and Modbus TCP “have a commanding lead in the number of nodes shipped” That is true as a percentage of the Industrial Ethernet market in the 2005 ARC report. We’ll have to wait for the next market share report to see where this stands today.
  • “over 1,300,000 total ODVA nodes as of 2006” is an interesting claim considering that in April of 2007 they were touting the number as 1.125 million in a press release.
  • “over 430 products” is an even more interesting claim since a search of the online catalog at ODVA shows 83 products from 41 vendors. And of course many of those are generic Ethernet switches, cables, and connectors that work with PROFINET, too.

In the PROFINET column under “Established”:

  • Refers to Siemens H1, but PROFINET has nothing to do with H1.
  • Peer-to-peer was introduced in 2001 (not 2002). We call this Component Based Automation (CBA) and comparing this to the peer-to-peer in the Ethernet/IP column is like comparing apples and oranges.CBA is an easy-to-use horizontal integration tool that integrates PROFINET devices AND most other fieldbuses.
  • When PROFINET IO originally shipped there were just a few products (duh), but there are over 115 products in the PROFINET Product News brochure dated November of 2006.
  • “Shipped about 8% as many nodes as EtherNet/IP in 2004” Let’s look at percentages in 2006.From April 2005 to April 2006 according to press releases from ODVA at Hanover Fair, 125,000 nodes of EIP were sold. That is 3.7% of the number of PROFIBUS nodes that were sold in 2006. Ok, that’s an apples-to-oranges comparison too, but it demonstrates the difference in market size between traditional fieldbuses and those based on Industrial Ethernet. And highlights why we think it’s important to consider the ease with which PROFINET and PROFIBUS play together.
  • “We” claim 14 million nodes of Siemens’ PROFIeverything, they say. They also say that includes all Interbus nodes. Let’s get those numbers up-to-date: through April of 2007 we have sold 20,000,000 nodes of just PROFIBUS. We know because most PROFIBUS devices use an ASIC. The ASIC is made in about 4 chip foundries. We count the chips; there are 20,000,000. I’m sure Siemens has a significant number of those, but by no means all. By the way, Interbus claims over 10,000,000 through 2006. The significance of Interbus is that this traditional fieldbus decided not to create an Ethernet version, but to use PROFINET instead. This means that the consortiums behind 30,000,000 installed nodes have chosen PROFINET as their Industrial Ethernet. DeviceNet does not publish their numbers, but reading Rockwell brochures and other public sources I’m guessing they might have 6,000,000 or about 20% of the combined PROFIBUS and Interbus base.
  • “We” have not released PROFINET totals. That’s true. We are working on that though. It’s not as easy as counting PROFIBUS chips since PROFINET works with standard Ethernet and does not require an ASIC. We have adopted a notary or independent auditor approach to poll the many PROFINET vendors and expect to be able to report numbers in 2008.

Moving on to the second row, “Standard”.

  • EIP “Works with standard commercial or industrial grade switches” as long as the switches have the “standard” IGMP option available only in more-expensive managed switches. Otherwise the entire network is flooded with EIP IO multicast messages. And, of course, if you plan to do motion control the switch must have the “standard” IEEE1588 option.
  • PROFINET also works with standard commercial or industrial switches but does not need the IGMP option. In fact, you can use an inexpensive unmanaged switch for PROFINET. If you need motion control with PROFINET, you need to implement hardware bandwidth reservation; the switch function, however, is unchanged. And it is not proprietary! The functionality is part of the open spec. Chips incorporating the bandwidth reservation system (plus a switch plus an Ethernet Controller – you need those functions anyway) are available from multiple suppliers including NEC and Hilscher.
  • EIP and PROFINET Ethernet frames are just Ethernet frames. Both are standard and can be viewed with an Ethernet sniffer like Wireshark. (Wireshark also decodes the open PROFINET protocol so users can see what the data in the frame means.)
  • PROFINET does not modify the OSI model!We do skip some layers, just not the same layers EIP skips.

The first page concludes with the section”More than a Fieldbus”:

  • The statements in the EIP column are probably more true of PROFINET than EIP! PROFINET works with standard switches. (PROFINET does not require IGMP which IT may or may not be familiar with.) PROFINET uses the exact same tools that IT already uses. PROFINET does not tunnel TCP/UDP/IP. PROFINET uses TCP/UDP/IP, just not for everything. Bypassing (not tunneling) TCP/UDP/IP allows PROFINET to be more deterministic and more repeatable than EIP according to tests conducted by the University of Michigan and reported in the Industrial Ethernet Book. There is nothing unknown about the performance; read the article “Technical Article: Performance metrics for Industrial Ethernet”.  We are sometimes asked why a target of 8 milliseconds was chosen for the test. I do know that PROFINET can achieve one millisecond (or better) for this kind of test. Can EIP? Anyway, PROFINET showed a consistent 100 microseconds of jitter; EIP had a significant number of messages arrive in 16 milliseconds instead of 8 (i.e., over 8 milliseconds of jitter).
  • PROFINET is PROFINET. It does three different things just like EIP does a couple different things (like motion control). Oh wait, EIP doesn’t have any shipping products that use EIP as a motion control bus; they just connect to a drive controller using EIP. PROFINET motion control products have been shipping for three years from multiple vendors.

One Response to “FUD Dissection Part 1”

« Back to PROFIblog