C Support of the CI and HSC Hardware

The Computer Interconnect (CI) bus is a high-speed, dual-path bus that connects processors and Hierarchical Storage Controllers (HSCs) in a computer room environment. An HSC is an I/O subsystem that is a self-contained, intelligent mass storage controller that provides access to disks and tapes from multiple host nodes attached to the CI bus.

Note
The Digital UNIX implementation has the following limitations:

You can attach a maximum of four HSCs to a CI bus.

You can attach a single CI bus to a host.

Under no circumstances can a Digital UNIX node participate as a VMS cluster member. A configuration that includes a VMS system and a Digital UNIX system residing on the same CI bus is not supported.

Digital UNIX supports Digital's System Communication Architecture (SCA) for CI port adapters and HSCs. SCA implements port and class driver support, and standardizes the ways in which TMSCP (tms) and MSCP (ra) devices are handled. SCA separates features into different architectural layers, thus minimizing the effect that software changes to one layer have on other layers.

C.1 Hardware Setup, Restrictions, and Revision Levels

For information on physical components and setup, refer to the HSC hardware documentation and the hardware documentation for your processor and supported devices. Only processors with CI adapters can support HSC configurations.

When setting up the HSC controller hardware, you should attach a terminal to the HSC in order to use commands to get or set HSC parameters, to monitor connections between remote systems, and to identify the disk or tape status.

The maximum number of hosts on a CI bus is 16. The host number for any host on the CI bus must be between 0 and 15.

Note
Two parameters of particular importance are the system ID and the system name. Do not duplicate any system identification or names of nodes on the star coupler.

C.2 Software Installation and Restrictions

The installation software assists you in identifying and configuring the components of your system. You should be familiar with the basic installation guide for your processor before starting the actual installation.

During installation of the Digital UNIX software, each accessible MSCP (ra) disk device must be uniquely identified by its unit plug number as follows:

The unit plug number must be between 0 and 254, inclusive.
Each unit plug number must be unique. Two different disks cannot have the same unit plug number even if the disks are on separate controllers. For example, if the unit plug number for a disk on controller A is 5 and the unit plug number for a separate disk on controller B is also 5, you must change one of the numbers.
You can connect a disk with a unique unit plug number to two different controllers (dual or porting). Refer to the ra(7) reference page for information on how to specify the device entry in the system configuration file.

C.3 Configuration File Entries

The installation software ensures that your HSC components are configured into the kernel and are included in the /usr/sys/conf/NAME system configuration file, where NAME specifies your system name in uppercase letters.

Chapter 5 provides information on the following entries that correspond to a CI or HSC configuration:

Description of the scs_sysid parameter
CI adapter specifications
Controller and device specifications

C.4 Booting an HSC Controller or an HSC Disk

The Digital UNIX software supports booting an HSC disk on the DEC 7000 and DEC 10000 processors. If an HSC controller fails, any disks connected to that HSC controller are inaccessible. Attempts to access those disks will cause the accessing system to hang until the HSC reboots completely. Refer to your processor hardware documentation for explicit instructions on booting an HSC disk.

C.5 Sharing Disk and Tape Units Among Several Hosts

Although an HSC can be shared among several hosts, there is no software interlocking mechanism to prevent concurrent write operations to the same partition by multiple Digital UNIX systems. The following restrictions must be observed:

Only multiple readers can share a disk unit; writable file systems cannot be shared.
If a disk will be shared, it should be hardware write protected.
Each host must mount the file systems to be accessed with the read-only (-r) option to the mount command.
Only a single host can mount a disk that contains writable file systems.
Use the Network File System (NFS) if multiple writers need to share partitions.

You should coordinate disk unit ownership among the hosts on the CI bus, for example, assign a range of disk unit numbers to each host. The HSC controller can also be directed to limit disk access to an exclusive host system. This limitation protects the disk from accidental access by another host on the CI bus. For more information, see the radisk(8) reference page, in particular the -e and -n options.

Tape drives that are attached to an HSC controller can be shared. This feature is recommended and provides greater use of tape drives. Be aware that the access mechanism provides serial sharing of the drives, not simultaneous access.