Unclustered TPC-C Results - another name for snake oil
By Todd Little-Oracle on Oct 14, 2009
One interesting thing to consider when looking at TPC-C results is to see if the results are for a clustered system or not. Non-clustered entries have only a single server performing the benchmark. If this server fails, the application is no longer available.
Now I'm sure there are customers running critical enterprise applications on non-clustered systems, but only those that enjoy downtime. It's virtually impossible to achieve the high availability that is expected of today's enterprise applications using only a single server. So when looking at performance benchmarks, make sure the environment you are evaluating can provide the level of availability you require. Otherwise you may be the hare instead of the tortoise (although Tuxedo proves you can have your cake and eat it too, just to mix my metaphors). Get the best Reliability, Availability, Scalability, and Performance you can for your C, C++, and COBOL SOA applications by running them on Tuxedo. Don't settle for systems with single points of failure.