What Decades-Old Environment Permits Parallel-Naive Developers to Keep a Large Parallel System Usefully Busy?

The answer is Structured Query Language (SQL), which has been used in production for decades in high-end datacenters. Yes, SQL is specialized. Yes, SQL is not perfect, as can be confirmed by asking Google about “SQL performance.”

However, it is a devastating counter-example to the all-too-common assertion that common applications simply cannot handle parallel machines. It can also serve as a nice counter-example to the also-common assertion that current software and practice is completely unequal to the task of creating parallel software artifacts.

In short: performance, productivity, generality. Pick any two. If you pick generality and performance, don't come running to me complaining about parallel programming being hard. Your own design decisions are after all what made it hard for your particular environment!

The designers of SQL made a different set of design decisions. I believe that the results speak for themselves.

Of course, there are other software environments with similar properties, particularly in the HPC area. However, these do not yet have SQL's decades-long record of heavy use in production environments.