A proposed SCADA system is to use 2x Sun Netra T1 sparc processors as Servers over 15 RTU's in a duty/ standby mode with 2 x operator workstations for scada control all on an Ethernet network.

What is the advantage of the extra hardware (sparc processor servers) when the 2 x operator workstations can be ran as servers?

 

The question is, do the (presumably existing?) workstations have the grunt and the stability to double up as the servers?

If they're too weak or already somewhat unstable, there's a case for buying the new machines and either adding them as servers, or using them to upgrade the operator stations to server status.

The advantage of separate servers is that if an operator station goes down, the whole system can continue (though you'll have to check what would happen in practice - presumably that operator station is there for a reason). Rebooting an operator station is also usually less disruptive than having to do the same to a server, though less so with dual ones. Depending on the situation, it might also be an advantage to have the servers locked and sealed where people can't get at them.

The disadvantage is that there's that much more stuff to go wrong.

Jiri
--
Jiri Baum <[email protected]> http://www.csse.monash.edu.au/~jirib
MAT LinuxPLC project --- http://mat.sf.net --- Machine Automation Tools

 

You need to look at a couple things to make a proper decision here.

1. Where are the operator nodes physically located in relation to the control system and server nodes? Typically, the servers will be located close to the controller, and will be physically connected to it. This architecture allows a lot of flexibility in adding operator nodes via ethernet. They can be anywhere you can run ethernet.

2. Will the system be expanded? Typically, a client/server architecture is implemented when there are several operator nodes in use, and having them all ask the controller for info is not practical. Having redundant servers allows only the servers to communicate with the control system, therefore making data flow much more efficient.

3. Although not critical, alarming can be an issue here as well. Depending on the syatem, alarms may be reported at the workstation or the controller. Alarms at the workstation may require acknowledging alarms at both workstations. Usually this can be avoided by some scripting, so it's not really a big deal, but still something to consider.

I have had a system that started out with one workstation talking directly to the controller. I expanded the system to redundant servers with six workstations. I do not advise planning a system to go from single workstation to redundant servers. There is an awful lot of work involved in re-referencing the points. Know what you want to do up front and where you expect the system to be in the next few years, and plan accordingly.

 

Thanks Jiri/ Dan

Both answer my queries. Infact a little bird tells me the system will expand to include 2 more trains if the project goes well.

Best regards & Happy New Year
Gray