User`s guide

ManualsBrandsDigital Equipment Corporation ManualsHome building and DecorStorage Works

2.1.3 CI, DSSI, or SCSI–2 Interfaces (Host Ports)

A CI, DSSI, or SCSI–2 interface allows direct memory access of data between the

host port and shared memory. Setup and maintenance of the host port is done by

the policy processor hardware.

2.1.4 SCSI–2 Device Ports (Buses)

The SCSI–2 device ports for the controllers are implemented using SCSI–2 port

processor chips performing 8-bit operations in normal or FAST mode. The port

processors execute scripts read from shared memory and under control of the

policy processor. Each SCSI–2 port can have up to six (or seven)

SCSI–2 devices.

2.1.5 Read Cache Module

The read cache used with HS array controllers reduces the controller’s latency

and may increase I/O’s per second. During normal operations, a read operation

from the host either accesses data currently in the controller’s cache or obtains

the requested data from a disk.

If a host read operation is a cache ‘‘hit’’ (data is already contained in the read

cache), the data is supplied back to the host immediately, thereby improving the

I/O performance by reducing latency.

If the host read operation is a cache ‘‘miss’’ (data is not in the read cache), the

controller accesses the appropriate disk to satisfy the request. The controller then

reads the data, returns it to the host, and writes it to the cache.

For host write operations, data is written to the cache and the disk. This write-

through caching improves the performance of subsequent reads of data previously

written.

Cache transfer sizes have a maximum size of 64 KB (128 logical blocks). Read

caching is enabled by default with 16 KB (32 logical blocks maximum transfer

size), and can be optionally enabled or disabled using the CLI SET unit-number

command. These cache parameters can be varied on a per unit basis.

Transfers that are larger than the maximum size (or 64 KB, whichever is smaller)

are not cached. This prevents large transfers from ﬂushing the cache.

The replacement algorithm implemented is a basic least recently used (LRU)

replacement algorithm. When the cache is full and new data must be written to

the cache, the LRU algorithm removes the oldest resident cached data with the

least frequent references and replaces it with the new data.

2.1.6 Write-Back Cache Module

The write-back cache used with HS array controllers may increase subsystem

performance, and it preserves data integrity under power failure situations.

For write-back caching, data is not always written to storage and cache

simultaneously, as it is for write-through caching. Instead, data intended for

storage may remain in the cache until the optimum time to write, or ﬂush,to

a device occurs. When data is held in this way it is referred to as unwritten

cached data. A power failure in conjunction with unwritten cached data can

have disastrous consequences if the information is lost. In RAID conﬁgurations,

A dual-redundant 6-port controller conﬁguration supports 36 devices (a 3-port dual-

redundant controller supports 18 devices). Lower availability conﬁgurations can support

up to 42 devices (7 per SCSI–2 port) with a 6-port controller, and up 21 devices with

a 3-port controller, but this will sacriﬁce a convenient upgrade to any of the possible

higher availability options, such as redundant power.

Controller Technical Description 2–3