Client Latency across Gateway

Ignition
,
1

Good afternoon,

I'm having issues with very slow clients that are currently connected to same gateway. Running 7.9. Gateway performance looks good, No memory or CPU issues. Its a fairly large line but similar in comparison to other lines that aren't having any issues. I had this issues recently which was connected to network latency. We have since fixed the network issue but these particular clients are still extremely slow. I've checked for sleeps, blocking, and run scripts. I've disabled them which didn't improve performance. the scan classes are not great but I have added classes and spread out the load and reduce the polling. They are set to read as was recommended in a previous post. Essentially tags are not being updated quickly even ones with 1s scan classes. All gateways are more or less the same from a design perspective.

Thanks!

2

Tag update performance is almost always PLC limited, not gateway limited.

Share more about you systems and comms layout.

3

The gateway is connected to 3 L84 Processors with an EN2TR as Host for the ignition gateway. All running on the one project. 21 Clients. All devices connect to there dedicated Stratix 8000 switch in each Panel. With the GB ports allocated for uplink and MQTT Server. Running 7.9.19 on Ignition and Java 8.

4

Each L84 has its own EN2TR.

5

Notes:

  • You tagged this as a Perspective question, but Perspective doesn't exist in v7.9.

  • Small differences in PLC comms configuration can make significant differences in performance, especially when other applications also talk to the same PLCs.

  • You should look at the Class3 messaging workload in the PLC's diagnostics. (Often reachable via the PLC's native webpage, in enabled in its project.) Compare that with PLCs that do not exhibit the problem.

2 Likes
6

What does the device status page on your gateway show for overload % on the devices with poor response times? This ties into pturmels' comments above with respect to PLC comms performance.

1 Like
7

I am currently running V31 on the PLC so I am unable to reference the Processor. One of the issues we had was the concurrent request was set to high. I changed the setting from Read to subscribed which has lowered the time and load for each scan class. But not have issues with the tags switching over and causing tag quality issues on screen. Highest load under device is 47%. Concurrent Request is set to 2 and CIP size is 500.

8

With the way Ignition's driver reports stats, this means your device is at 147%. Any percentage over 0% means your device is lagging behind. Missed the 7.9 tag

Are your tags coming from all over on the device or are they organized to be in contiguous sections of memory?

You can try bumping the CIP size to something like 2000 or 4000 to reduce request counts, but if your requested tags are not near each other in memory you will still end up with a higher request count.

Approximately how many tags are you pulling from each device?

9

Not true in v7.9. The change from "Load" metrics to "Overload" metrics didn't occur until v8.1.7.

1 Like
10

140k Tags total. Roughly the same tag count for each device. Were running the same project on each client. Some clients are configured for certain stations with higher tag subscriptions but for the most part all clients are slow.

11

Reviewing your OP further, I suspect that your problem is OS-specific on the troublesome client PCs, if not still network latency. (Running Vision clients over a slow VPN will cause any bad design/scripting practices to blow up in your face.)

12

Each client is its own physical windows terminal connected via ethernet to the Stratix 8000 switch of each device. All clients in department are the same with other gateways that have larger tag counts but only difference being running a project per device. I'm not sure how ignition manages more projects via tags. Although, smaller tag count per project.