Browse Prior Art Database

ASCM Module Bus I/O Controller

IP.com Disclosure Number: IPCOM000113066D
Original Publication Date: 1994-Jul-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 4 page(s) / 101K

Publishing Venue

IBM

Related People

Langston, DG: AUTHOR

Abstract

A controller for attaching I/O devices directly to the ASCM module bus without the presence of a local processor is disclosed. The I/O controller uses a priority token passing scheme for selecting from up to seven I/O devices.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

ASCM Module Bus I/O Controller

      A controller for attaching I/O devices directly to the ASCM
module bus without the presence of a local processor is disclosed.
The I/O controller uses a priority token passing scheme for selecting
from up to seven I/O devices.

      The I/O controller architecture is a very simple architecture
being derived from the DCE structure of the X.25 protocol.  At its
simplest, the I/O controller can be described as a finite queue used
to multiplex several I/O protocol chips with the ASCM module bus.
The high level dataflow architecture is shown in Fig. 1.

      The I/O bus is a token bus that is polled using a prioritized
roll call polling scheme.  The polling scheme is in effect four
schemes rolled into one.  The first poll priority is the 1-2-4
scheme.  Here one device gets four polls per polling cycle, two
devices get two polls each and four devices get one poll each.  This
allows for three levels of I/O performance.  The second poll priority
is the 3-4 scheme; here, three devices get three polls  and four
devices get one poll per cycle.  The third scheme is standard roll
call polling where every device is polled once per cycle.  The final
and highest throughput scheme is the 1-1-1-4 scheme where one device
gets 4 polls, the next gets three, the next two, and the rest one
poll.

      The polling cycle is defined here as the time it takes to
completely poll all of the devices.  To complete the 1-2-4 scheme 12
polling actions are required.  To complete the 3-4 scheme and 1-1-1-4
scheme 13 polling actions are required.  The standard roll call
polling scheme requires 7 actions.  In addition to data transfer,
control instructions to the processor from the I/O devices and vice
versa are required.  To allow for this dead cycles will be placed in
the polling cycle to allow these transfers.  Therefore, 16
pseudo-polling actions will be required to completely poll the
maximum number of I/O devices which is 7.

      These polling schemes work well for either duplex or half
duplex operation.  To allow maximum throughput, the I/O controller is
designed to have full duplex capa...