As of Version 7.0, OpenVMS Alpha includes two features thatprovide dramatically improved I/O performance: Fast I/O and FastPath. These features are designed to promote OpenVMS as a leadingplatform for database systems. Performance improvement results fromreducing the CPU cost per I/O request and improving symmetric multiprocessing(SMP) scaling of I/O operations. The CPU cost per I/O is reducedby optimizing code for high-volume I/O and by using better SMP CPUmemory cache. SMP scaling of I/O is increased by reducing the numberof spinlocks taken per I/O and by substituting finer-granularityspinlocks for global spinlocks.
The improvements follow a natural division that already existsbetween the device-independent and device-dependent layers in theOpenVMS I/O subsystem. The device-independent overhead is addressedby Fast I/O, which is a set of lean system services that can substitutefor certain $QIO operations. Using these services requires somecoding changes in existing applications, but the changes are usuallymodest and well contained. The device-dependent overhead is addressedby Fast Path, which is an optional performance feature that createsa "fast path" to the device. It requires no applicationchanges.
Fast I/O and Fast Path can be used independently; however,together they can provide a 45 percent reduction in CPU cost perI/O on uniprocessor systems and a 52 percent reduction on multiprocessorsystems.
HP OpenVMS I/O User's Reference Manual
