CPU313C+CP343-1 Communication lag while both OPC and HMI are accessing

[arthur.hauer]

Dear experts,

I have the following setup:

CPU-313C (6ES7313-5BF03-0AB0 V2.6) + CP343-1 (6GK7343-1EX30-0XE0 V2.0) communicating with Siemens PC HMI (not sure what protocol it's using) and Kepware OPC server (TCP/IP) via RJ-45.

When both OPC and HMI are acessing data (1~2s window between two data access requests), the HMI interface gets really 'laggy', reaching up to 30~40s between bit switch command(push of an on-screen button, for example) and actual action. The same behaviour goes for OPC server, although, commands are not as delayed as on the HMI.

I've tried reserving more connection resources for PG communication to no avail. Also tried increasing max scan cycle load from communication (from 20% to 50%), also to no avail.

Scan time didn't change (~35ms) when both were connected.

On CP 343-1 configurations, both ports are set to 'automatic setting'.

I've also observed the same behaviour when monitoring DBs.

Does anyone have any solution for this problem (or, at least, an explanation for why it happens)?

Att.,
Arthur

[JesperMP]

All the data has to pass over the backplane of the S7-300 which is the slow MPI bus @187.5kbps.
The combination of the HMI and OPC reads or writes too much data too often.
That the PLC has 35 ms scan time also tells us that the CPU is at the limit. That does not help either.

Proper remedy is to use a CPU that is suited for the job. Either a 315-2PN/DP or a 317-2PN/DP. These are more powerful and has an integrated ethernet port that is much much more powerful than MPI.

A short term remedy is to modify the data.
The usual methods:
Cut out all data that is not really necessary.
Increase the update times generally.
Avoid 100ms.
Let the fastest be 500 ms, and chose greater if possible. Temperatures changes slowly, so these can be 10 s for example.
And group similar data together, dont read or write randomly.

[arthur.hauer]

Originally Posted by JesperMP

All the data has to pass over the backplane of the S7-300 which is the slow MPI bus @187.5kbps.
The combination of the HMI and OPC reads or writes too much data too often.
That the PLC has 35 ms scan time also tells us that the CPU is at the limit. That does not help either.

Proper remedy is to use a CPU that is suited for the job. Either a 315-2PN/DP or a 317-2PN/DP. These are more powerful and has an integrated ethernet port that is much much more powerful than MPI.

A short term remedy is to modify the data.
The usual methods:
Cut out all data that is not really necessary.
Increase the update times generally.
Avoid 100ms.
Let the fastest be 500 ms, and chose greater if possible. Temperatures changes slowly, so these can be 10 s for example.
And group similar data together, dont read or write randomly.

Jesper,

That was my first guess, but wanted a second opinion.
I honestly can't remember if I've already used an equal (or inferior) CPU in a similar arrangement.
One thing that got me scratching my head is the fact that we have S5s running smoothly with the 'OPC+HMI' setup (with an MPI to TCP/IP adapter).

Just for fun, a quick comparisson of response times between the mentioned CPU313 setup and a 1200 CPU (measured with Kepware OPC):

CPU313C - ~210ms
1200 - ~20ms

Thanks for the help!

[arthur.hauer]

Originally Posted by JesperMP

All the data has to pass over the backplane of the S7-300 which is the slow MPI bus @187.5kbps.
The combination of the HMI and OPC reads or writes too much data too often.
That the PLC has 35 ms scan time also tells us that the CPU is at the limit. That does not help either.

Proper remedy is to use a CPU that is suited for the job. Either a 315-2PN/DP or a 317-2PN/DP. These are more powerful and has an integrated ethernet port that is much much more powerful than MPI.

A short term remedy is to modify the data.
The usual methods:
Cut out all data that is not really necessary.
Increase the update times generally.
Avoid 100ms.
Let the fastest be 500 ms, and chose greater if possible. Temperatures changes slowly, so these can be 10 s for example.
And group similar data together, dont read or write randomly.

Oh, by the way:
your suggestions are completely valid, but, for our data access requirements, not enough...
I'll sugest a CPU upgrade.

Again, thanks for the help!

[Lare]

Originally Posted by arthur.hauer

Jesper,

That was my first guess, but wanted a second opinion.
I honestly can't remember if I've already used an equal (or inferior) CPU in a similar arrangement.
One thing that got me scratching my head is the fact that we have S5s running smoothly with the 'OPC+HMI' setup (with an MPI to TCP/IP adapter).

Is tags updated also on S5s CPU as fast as tags are readed. I mean is there buffer on MPI adapter for tags and it looks fast on HMI but actually is slower as tags are readed from adapters buffer not directly from CPU tags.

[JesperMP]

Originally Posted by arthur.hauer

One thing that got me scratching my head is the fact that we have S5s running smoothly with the 'OPC+HMI' setup (with an MPI to TCP/IP adapter).

Since you menation an MPI to TCP/IP adapter, you probably mean some kind of Profibus to TCP/IP adapter (*). MPI was newer supported on S5, but Profibus was. And a Profibus CP in an S5-115U or S5-135U rack can be reasonably fast.


*: or an adapter that supports but MPI and Profibus.

[arthur.hauer]

Originally Posted by JesperMP

Since you menation an MPI to TCP/IP adapter, you probably mean some kind of Profibus to TCP/IP adapter (*). MPI was newer supported on S5, but Profibus was. And a Profibus CP in an S5-115U or S5-135U rack can be reasonably fast.


*: or an adapter that supports but MPI and Profibus.

My mistake... You are correct.
It's a TCP/AS511 adapter (https://www.process-informatik.de/93...html/?lang=en_)

[JesperMP]

Originally Posted by arthur.hauer

My mistake... You are correct.
It's a TCP/AS511 adapter (https://www.process-informatik.de/93...html/?lang=en_)

That is an adapter for the S5 PG port, not MPI or DP.
In that case I am certain that Lare is correct, despite that the data seems to update fast, in reality is merely the data that is buffered in the adapter. The real data transfer is much much slower.
The PG port in an S5 PLC (with AS511 protocol) operates at 9600 baud (!).
So approximately 1/20 the bandwith of MPI @ 187.5 kbps, or 1/10000 the bandwidth of Ethenet @ 100 Mbps.