I/O Express uses available host system memory to store actively accessed data blocks. This allows subsequent requests for the same blocks to be satisfied from memory, instead of from disk. I/O Express is unique because it is self-monitoring, dynamic, application-independent, and fully functional on all OpenVMS configurations. It is effective and easy!
I/O Express dynamically allocates and deallocates system memory for the data caching operation. It constantly evaluates the available memory on the host system. When lots of memory is available, I/O Express allows itself to hold a large amount of disk data in memory. When the amount of free memory decreases, I/O Express returns memory to the OpenVMS operating system. So as the memory requirements of other processes fluctuate, the resources used by I/O Express are adjusted to accommodate system needs.
I/O Express is fully automatic! There is no need to predict which files to cache. Selection of disk blocks to hold in memory is based on a sophisticated caching algorithm. Large files are not stored in memory-based cache in their entirety, only those blocks likely to be read soon.
I/O Express is a write-through cache. Both read and write operations are handled efficiently by I/O Express. On a typical system, a high percentage of read I/0 operations will be satisfied from the memory-resident data cache. At the same time any write or update of the cashed data is instantly written to the disk as well. So the cached data is invulnerable to system failure.
I/O Express communicates smoothly between nodes of OpenVMS clusters. This includes all types of clusters. As with stand-alone systems, there are no restrictions regarding files or disks that may be cached on an OpenVMS cluster.
Best of all, I/O Express accomplishes this at practically no cost in system resources!
The memory used by I/O Express would not otherwise be used. The system's capacity for 1/0 throughput increases dramatically. I/O Express does more than reduce data access time. CPU overhead required to satisfy an I/0 request from I/O Express's memory-resident cache is often less than CPU overhead would be without I/O Express!
|