There could be a number of answers to your question, depending on where you look at it from. Here's what I learned from my experience: We were to redesign the PLC code for an already-running concrete-brick-making plant, as part of a preemptive disaster recovery plan. It was a big plant with two major control rooms, one for the brick-making link, and one for batching/weighing and mix preparation. Both rooms had a panel with PLC racks, containing I/O modules, analog modules, etc. However, the whole plant was controlled by one PLC. The panel in the batching control room only had I/O and analog modules, connected by an interface module to the main PLC, as an extension to the main rack.

While the system was working fine under the control of one PLC, I see this as an example of cost-cutting. When the PLC finally gave out, taking the plant down with it, the damage was done. There was a temporary shutdown in the brick-making section. But the real damage was done in the batch weighing and mixing section. There was wasted material and extra labor charges to practically blast the set mixture out of the mixers/hoppers.

Off-the-shelf products have become a very cost-effective way to tackle big projects. It's easy to solve an issue like the one Zeeshan encountered by having redundancy for the processor and I/O.

The factors that should be considered when deciding whether to split a project over multiple CPUs are:

• Independence, or the ability to completely shut down one part of the plant without affecting other parts.

• Cycle time. If your program reaches 80 percent of the design cycle time, you should look for an extra controller. However, if you decide to split the program at this point, the cost will probably be much higher than if you use two controllers right from the get-go.

• Memory. Similar argument -- if you use up 80 percent of the available memory, you need to consider splitting.

• I/O count. If you run out of addressing space or possible extension space, you'll be forced to split. However, this usually occurs early in the design.

• Distance. It may be cheaper to add another controller at another location rather than pulling all the extra lengths of cable. However, this argument is fading as most PLC systems use remote I/O over profibus or profinet.