HP OpenVMS Alpha Version 8.3 and HP OpenVMS Version 8.3-1H1 for Integrity Servers SPD 82.35.13

Memory errors. These can be unrecoverable (hard) errors or recoverable (soft) errors. The system ex- amines memory at startup time and does not use any bad pages. During system operation, the system cor- rects all single-bit memory errors for those systems with error correction code (ECC) memory.

Correctible memory errors. A primary cause of these correctible memory errors is alpha particle radi- ation. On some processors, when correctible mem- ory errors occur, the memory controller corrects only the data returned to the CPU or I/O controller. The actual data in memory is left with the error intact. Subsequent read operations cause correction cycles to occur and, in most cases, an interrupt to report the error. On many of these processors, OpenVMS monitors the occurrence of correctible memory errors and, in almost all cases, is able to remove the error condition by rewriting the data in memory. Rewrit- ing the data causes the data to be corrected in that memory location.

Other failures include:

Operating system errors (system-detected inconsis- tencies or architectural errors in system context)

User errors

I/O errors

The system logs all processor errors, all operating system errors detected through internal consistency checks, all double-bit memory errors (and a summary of corrected single-bit memory errors), and most I/O er- rors.

If the system is shut down because of an unrecoverable hardware or software error, a dump of physical memory is written. The dump includes the contents of the processor registers. The OpenVMS System Dump Analyzer (SDA) utility is provided for analyzing memory dumps.

Input/Output

The QIO system service and other related I/O services provide a direct interface to the operating system’s I/O routines. These services are available from within most OpenVMS programming languages and can be used to perform low-level I/O operations efficiently with a minimal amount of system overhead for time-critical appli- cations.

Device drivers execute I/O instructions to transfer data to and from a device and to communicate directly with an I/O device. Each type of I/O device requires its own driver. HP supplies drivers for all devices supported by the OpenVMS operating system and provides QIO system service routines to access the special features available in many of these devices.

OpenVMS supports a variety of disk and tape peripheral devices, as well as terminals, networks, and mailboxes (virtual devices for interprocess communication), and more general I/O devices.

I/O Performance Features

Fast I/O provides a suite of additional system services that applications can use to improve I/O throughput. The fast I/O services minimize the CPU resources required to perform I/O.

Fast Path provides a streamlined mainline code path through the I/O subsystem to improve both uniprocessor and multiprocessor I/O performance. On multiprocessor systems, Fast Path allows all CPU processing for specific I/O adapters to be handled by a specific CPU. This can significantly lower the demands on the primary CPU and increase the I/O throughput on multiprocessor systems with multiple I/O ports. No user application changes are needed to take advantage of Fast Path. Fast Path can be utilized by the $QIO system service or the Fast I/O services.

Extended File Cache (XFC)

The Extended File Cache (XFC) is a virtual block data cache provided with OpenVMS Alpha and OpenVMS for Integrity servers. Similar to the Virtual I/O Cache, the XFC is a clusterwide, file system data cache. Both file system data caches are compatible and coexist in the OpenVMS Cluster.

The XFC improves I/O performance with the following features that are not available with the virtual I/O cache:

Read-ahead caching

Automatic resizing of the cache

Larger maximum cache size

No limit on the number of closed files that can be cached

Control over the maximum size of I/O that can be cached

Control over whether cache memory is static or dy- namic

Virtual I/O Cache (Alpha only)

OpenVMS Alpha provides a standalone or clusterwide, file-oriented disk cache. Applications benefit from the advantages of the virtual I/O cache without any special coding. The virtual I/O file-caching algorithm is chosen based on the type of clusterwide access currently in progress. Virtual I/O caching reduces current and potential I/O bottlenecks within OpenVMS systems. It reduces the number of I/Os to the disk subsystem, thereby reducing systemwide bottlenecks.

13