Distributed Focussed Availability

Distributed and Redundant Systems Architectures

Any automation system component may fail, usually at an inconvenient time. Whenever a failure happens, some part of the operation will be affected.  What and for how long depends on the way in which the automation system is arranged.  This post covers two methods of arrangement.  Distributed on the one hand, where the automation functionality required for different parts of the plant is distributed with the plant. Redundant on the other hand, which centralizes functionality and provides high availability through the provision of a duplicate backup system. Invensys provides both redundant and distributed RTU types to cover many different forms of remote automation architecture.

Lets consider each type of arrangement.  For the distributed system, cells of autonomous functionality are arranged and connected by a network.

Network

The most important aspect of this is that although any node [1..N per the fig.] can fail, it will not fail it's network, nor will it's own failure cause any other node to fail. For the redundant system, the main [A in the fig.] system component hosts all of the functionality, and should it fail, then the entire functionality will be taken over by the [B] system host.  A failure on the network between the hosts has to be guaranteed not to happen.

Nodes

Each node [numbered gold blocks in the fig.] on the distributed network is considered a functional host.  In case of necessity or programming, each host can talk to any other host over the network, usually in a token ring or deterministic way.  There is usually capability of distributing a large number of nodal functional hosts around the ring.

Each node in a redundant network [un-numbered white blocks in the fig.] has it's functionality removed and located in a central host [A,B in the fig.].  The nodal functionality then becomes either one of providing remote IO or being simple wiring back to the centralized location.

Functional Automation Hosts

A distributed NODE generally has no backup included.  Each node stands on it's own, and should it fail, then the functionality associated with that location becomes unavailable. Given that plant equipment designs include functional asset redundancy, and that SCADA functionality is supervisory in nature, the distributed model provides many functional and scalable advantages.

A redundant arrangement consisting of [A,B] hosts centralizes functionality from all the would be nodal locations and runs the automation of the A,B hosts in parallel, each host doing all of the data acquisition in parallel, but generally only one of the hosts issuing commands to the field.  The choice and determination of which of the [A,B] hosts should be in command is handled by interhost communications [black arrow in the fig.].

Why prefer one architecture over another ?

The application determines the best advantage of distributed vs redundant architectures.  For mission critical applications where all of the functionality is highly cross coordinated and all of it has to be made at the highest availability level, the redundant architecture is generally the most suitable.  Distributed systems have an advantage of being able to withstand one failure for whatever reason, and only impact that particular function.  Distributed systems are particularly suitable plants with multiple units or pathways, allowing for plant level redundancy to be assured regardless of the cause of the automation failure.

In the SCADA industry we find that distributed systems are very effective for substation automation where substations consist of multiple feeders and redundant pathways, and the outage of any particular feeder should NOT impact on any other one. Redundant systems are generally required for oil and gas stations where possibly only one pump or machine is available and the automation system availability has to be much higher than that of the plant.

For distributed architectures, our preferred offering is the SCD5200, which can be distributed up to 50 Nodes on a high speed networking ring.

For redundant architectures, our preferred offering is the SCD2200, which generally offers centralized IO and redundant communications and logic processing.