IBM System/370 Manual page 38

time. For the 3330 facility, for example, sector time is defined to
be approximately
130
microseconds. Thus, there are
128
sectors per
track on the
3330.
A disk control unit with RPS and multiple requesting can determine
the sector currently under the heads of each of its drives. A sector
counter is contained in each drive. It is incremented once every
sector time period and set to zero each time the index marker passes
under the heads. The sector in which a record falls is a function
of the length of all records that precede it and of its sequential
position on the track. Therefore, sector location can be calculated
for fixed-length records.
Two new disk commands are provided for use with rotational position
sensing:
SET SECTOR
READ SECTOR
If the sector address of a record is known or can be calculated,
a SET SECTOR command can
be
included in the disk channel program to
cause the control unit to look for the designated sector. Once the
control unit accepts the sector number provided by a SET SECTOR command,
both the block multiplexer channel and the disk control unit disconnect
and are available for another I/O operation. When SET SECTOR is used
for positioning, the time the channel is busy searching for a record
is reduced from an average of
8.3
ms to an average of
250
microseconds
for the
3330
facility. (Allowing for the worst case of speed variation
and for disk pack interchange, the search time for a record, from
sector found to beginning of desired record, can vary from
120
to
380
microseconds on a 3330 facility.)
The READ SECTOR command is useful for sequential disk processing
and for write verification. When chained from a READ, WRITE, or SEARCH
command, READ SECTOR provides the sector number required to access
the record processed by the previous CCW. This sector number can be
used to reposition the track to the record in order to verify the
record just written or in order to read or write the next sequential
record. These two new sector commands, used in conjunction with the
block multiplexer channel, permit a single command-chained channel
program to be initiated for each disk operation to free the channel
and disk control unit during' seek and rotational positioning operations.
When record 10 is known, the two channel programs shown below can
be used to retrieve a record directly from an OS BDAM data set on a
direct access device without RPS, such as the
2314
(key was not
written). The seek operation can be overlapped with other seeks and
one data transfer operation on the same selector channel.
Channel program
1.
Initiate the stand-alone seek to position the disk
arm.
Command Chaining
Flag Command
Selector Channel
and Disk Control
Unit Status
SEEK (Seek address)
Free as soon
as the control
unit accepts the
seek address
31