Control Logix Ethernet Troubleshooting

[cntrlfrk]

I am currently using a Control Logix Processor (L55) with an ethernet bridge (1756-ENBT/A) and accessing it with a half dozen HMI PC's, each with about 1000-1500 tags.

The problem is, sometimes the network seems really bogged down, and some data on the HMI's will not appear for some time. The plant network is 'isolated' with only the PLC and HMI's on one switch, and an uplink to the office switch for data collection. I've looked in the port diagnostics on the ENBT, and didn't really see anything that seemed out of the ordinary.

I'm wondering if I have one problem child that seems to cause problems for the whole network since this doesn't seem to be an overloaded network to me.

Does anyone have any ideas on how to isolate or troubleshoot this further??

 

[dmroeder]

There is a packet sniffing utility called Ethereal that will allow you to capture packtes and analyze data on your network.

Tell me about the switch? Would it happen to be a managed industrial switch? Apparently this is a must for a decent size network with Ethrnet/IP traffic.

Edit:
"They" say that the switch should also support IGMP snooping

 

[cntrlfrk]

The switch is a manageable switch, I believe, a 3COM 4500, but we basically took it out of the box and plugged it in, we haven't gone in and prioritized or done any network management.

 

[MartB]

What HMI software are you using ?

The reason I ask is that I had a similar situation about a year ago with RSView and 12 Versaviews. For no apparent reason the network would run slow (sometimes even to a standstill). We even had Rockwell come to site to investigate. Eventually finding out that RSView v4 and above had/have networking issues with XP SP2. We dropped down to version 3.x and all of our problems went away and we have had no network issues since then.

Just my 2c worth.

 

[dmroeder]

Now, I'm kind of new to this level of networking. I've always just plugged cables into switches and got them communicating. I have recently been trying to eductate myself about what is really going on and really trying to understand Ethernet communications.

From what I understand, a switch will learn where packets are coming and going and then direct them accordingly, to minimize collisions on the network. Apparently Ethernet/IP uses the UDP protocol for some of the communications, which broadcasts packets everywhere on the network, and the devices that need them accept them. This creates a lot of traffic on a network if there are a lot of UDP packets being sent. IGMP snooping is supposed to eleminate help organize these packets on the network.

You can use Ethereal to monitor what kinds of packets are on your network.

Also, you might look into using the QoS feature in your switch. This feature will allow you to prioritize packets on the network.


Again, I'm kind of a n00b (a Rook if you will) at this networking business, so hopefully my information was not misleading. It will at least give you some things to look at. If I have been misleading, I'm sure someone will correct me (hopefully)!

 

[rdrast]

If you have a chance, try swapping out the L55 CPU with an L6x series, and see if you notice a difference.

I had to do that on an installation with multiple PV+ units; The PLC Comms bogged down so much it was almost impossible to even get online with it. If the PV+'s could have read the tags out of Linx or Linx Enterprise on the SE Server, there would have been no problem.
As it is, each PV+ must create it's own comms link to the CPU. The L6x series handles this MUCH better than the L55 series.

 

[CroCop]

1. Enable IGMP snooping on the switch.
2. Increase the CLX's SOTS (right click the controller in the pane, click advanced, and up the System Overhead Time Slice). By upping this, you will afford more time for communications & less for logic processing.

 

[cntrlfrk]

These are all RSView32 machines, half Versa-View Touchscreens, the rest just PC's.

I downloaded Ethereal, I'll have to do some more with it but initially I didn't see how it anaylzed the network, it looked like it only analyzed the network adapter on the machine it is loaded on, but, like I said, I'll need to look into that some more.

I checked the 'Overhead Time Slice'. It is set at 10%. I noticed a CLX we have in another plant is set to 20%. I assume I can change this online?

I don't really have the option to swap controllers at this time but will keep that in mind for the future.

I am not currently at the plant, but will have to locate the software next time I am there to enable the IGMP snooping. I do have access to a computer there over the internet, I may be able to do it that way.

 

[CroCop]

Yes you can change the SOTS online. 10% is pretty low IMO.

 

[MartB]

Most of the systems I've installed recently have needed a STOS of around 40% + just to keep running without errors, but it really depends on how much communication is going on. I've has some systems that required 60% in the past.

I think the default is 20% (I could be wrong about that though), so yes I'd agree that 10% is very low.

 

[Contr_Conn]

accessing it with a half dozen HMI PC's, each with about 1000-1500 tags.

With 10% SOTS this will be definately issue.
Bump it up to 40% to see if it works any better, controller will have more time servicing comms.
Watch program scan time - it will definately go up unless your program has only periodic tasks.
IF scan time is issue, then upgrade to L6x controller.

Also watch ENBT CPU utilization, this may be your next bottleneck.

 

[Operaghost]

Most managed switches that I have seen operate as a standard switch if you don't configure them "out of the box". So that means it is not really doing what a managed switch can and should do.

So you will need to get into the switch and turn on those special features such as IGMP snooping and QoS. Also, when you get into it, you can monitor port usage and even the type of data coming and going. Check out the 3com web site as you can view specs and probably manuals as well.

OG