PEER TO PEER vs. SERVER-CLIENT ???

I AM RESEARCHING AN HMI UPGRADE FOR OUR PLANT. I AM GETTING MIXED OPINIONS ON WHAT IS BETTER ARCHITECTURE...PEER TO PEER OR SERVER-CLIENT. I WOULD APPRECIATE A FEW MORE OPINIONS ON THIS SUBJECT MATTER.

You may not have any choice - many protocols only support one architecture. What kind of PLC are you using, and what kind of network do you have now? How large is the network? Do you have lots of identical units, or only a few stations? Are you doing control on the network, or only data logging? Do you need fast response or is several seconds latency OK? Do you need PC connectivity or only HMI terminals? Do you have devices from several manufacturers or just one.

These kinds of questions will narrow your choices. If there were one "best" architecture for control it would dominate the industry. The best architecture for one application may not be best for another.

I generally prefer server/client (master/slave) because I am in control of all transactions, but in some projects peer to peer has been better.

Thanks Tom, you have really opened my eyes to the factors involved and I now realize how little about this I know. I will describe our network and perhaps you would be kind enough to comment further or anyone else for that matter.

I work at a cement plant...and am very new on the job. We have a Modicon 984 E PLCs running a modbus and Token Ring LAN. We are planning to upgrade to Modicon Quantum PLCs with an Ethernet LAN. There are a few thousand I/O points. There will be four to six operator stations and several monitoring stations. Each operator station is used to control different plant processes all occuring simutaneously. 24 hours-7 day week process. Data logging, trending, alarming, etc. will all be required of the new system. There will be a hand full of closed loop controlled equipment. Relatively fast response time would be a plus. The operator stations are PCs which will run an HMI software program such as Wonerware or something similar. Field devices are from many different vendors. Most important is not loosing control of the system. If control is lost of a cement kiln, the lively hood of the plant would be in serious jeopardy! While manual overide is possible, quality production is not which obviously results in serious money losses and problems. I would certainly appreciate more comments on the peer to peer vs server-client HMI architectural advantages for this type of plant. Thank you for your help.

Ethernet is inherrently peer to peer architecture, so that decision is made.

You can use a peer to peer architecture in what amounts to a master slave mode, simply by how you structure the data inquiries and who asks for data. I have done this using ethernet and GE 90-30 PLCs simply by having the master PLC initiate data inquiries to three identical slave units. The master did overall process and equipment coordination, and the slaves handled the individual machine control. A SCADA system was tied into the same network and obtained over all critical process data from the master. Each PLC, master and slave, had a HMI terminal using the serial connection to a second port on each PLC. That way the HMI response was fast, we could display all the data, and the bandwidth on the ethernet wasn't used up.

You can do this with the Quantum PLC as well. That way you can get fast response to each kiln's or crusher's HMI, the individual process control and protection isn't compromised, and each kiln's proces control is in local logic at that PLC. All the data on speed, loading, temperature, etc. can be collected by the PCs and trended and data logged. I suggest you contact your Modicon / Square D distributor and get them involved early on.

On a cement plant, you really cannot afford to mess around on the automation front. Basically because long kiln stops cost huge amounts of money as you are probably aware.
You need a system that is reliable 24x7 365 days of the year. Which ever you choose.

I would recommend splitting your kiln into departments to make the job easier and upgrade one department to your new system every time you stop the kiln for maintenance, brickwork or whatever.

Of course doing it like this will cost you more money than doing it all at once. Have you the time to do it all at once? Never mind the resources? You will probably need a programmer to work on each department.

Have you thought about getting an automation cement specialist in on the job? The money you spend could be returned by the job not being overrun. Or are your boots big enough for this one with your own team? Please think wisely on this one.