A great ‘What You Should Know About Server Clustering’ piece published by Dedicated Server School brings to light the problem with clustering, i.e. that it tries to do two things at once – a bit like the front wheels in a front-wheel drive car.
In fact, just like the front wheels in a front-wheel drive car, there comes a point when nothing works. That’s why F1 cars are always rear-wheel drive!
Clustering for High Availability is just too complex anyway and will lead to a management nightmare – eventually it will fail and cause downtime.
Clustering for improved performance is something different – the scale out approach can be used very effectively.
For the best of both worlds, combine this scale-out and run on fault tolerant hardware to protect against downtime.