This is a questions that will have to be answered by the folks responsible
for Perfect Cache.
Depending on the application, RMS buffering can be superior to any disk cache.
The best I/O is the one you never did. This is the idea behind both RMS
buffering and a disk cache. RMS buffering has the advantage in that it can
service an application request without going through the OpenVMS device driver
routines. Every disk cache is at the disadvantage in that the application
request goes through RMS first anyway. The sad thing is, even with both RMS
buffering and a disk cache, after rumaging through the in memory data, the
request might still have to go to a disk. If we had known, we could have
skipped directly to the physical I/O and saved time. Answer, get enough memory
to hold enough data to keep the physical I/O to a minimum. The nice thing is
the search through the RMS buffers and the disk cache scales with processor
speed. Faster processors can handle bigger memories. Processors are not
infinite in speed. They do have lots of things to do besides looking around for
data. If your RMS buffers and disk cache are too big then you, yes, you might
find the data in memory but you've tied up the processor when it could have been
off doing more useful work. After all, why did you invest in the mega-I/O
subsystem if you aren't going to give it some work to do in parallel with the
processor?
Yes, it is possible to write an application which doesn't use RMS and goes
directly to the QIO instead.
QIO has been optimized considerably in V7.0 under Fast Path and Fast I/O.
