The Electric Power Research Institute has published a very useful guide to what can go wrong with power stations due to y2k. It also suggests a plan of attack. It says that the fix can take as little as one year. Let's hope they're right. If it takes longer than two years, the entire civilization is at risk.
Embedded chips cannot be reprogrammed. They are everywhere in old systems. The industry has a lot of old systems. Here's why.
* * * * * * *
Real-time systems can be very complex, and they are used to control or monitor very high-value processes. Typically, a power station will have scores of real-time systems. They have been bought for different reasons by different people over the years, usually mirroring the gradual development of the installation. The production processes are now dependent on the successful continuous operation of the real-time systems.
Because the production processes are so valuable, production managers and engineering staff fear the failure of real-time systems. When real-time systems fail, high-value processes shut down, and the costs of unexpected shutdowns can be enormous. For a power station, the cost of an unexpected shutdown can be hundreds of thousands of dollars. The pressure to keep the production process running is great. As a result, production managers resist changes to embedded systems.
This means that operating systems are not upgraded. Improved functionality is postponed. Hardware which is no longer supported by the manufacturer remains in use.
The result is a bunch of ageing systems, based on languages, packages and processors for which the skills are gradually being lost. Because of this culture, fixing the Year2000 problems is more complicated than for banking or administrative applications. The systems are more difficult to audit, because some are so old that the information about them has been lost. Doing the triage is complicated, because there is a risk that taking the system through a mock millennium change will cause the process to fail, with huge cost penalties.
So to fix the problems, we need people who understand embedded systems technology, the production processes, and the commercial impact of mistakes in a manufacturing environment. These people are rare.
Systems are not yet failing, because real-time systems tend to have a lookahead of less than a month. So the failures will come late in 1999.