Browse Prior Art Database

Method And Apparatus for Data Flow Control On a Serial I/O Channel

IP.com Disclosure Number: IPCOM000120339D
Original Publication Date: 1991-Apr-01
Included in the Prior Art Database: 2005-Apr-02
Document File: 3 page(s) / 118K

Publishing Venue

IBM

Related People

Elliott, JC: AUTHOR [+6]

Abstract

In an I/O architecture based on transmission of data in packets or frames without byte-by-byte handshaking, a flow control or pacing procedure is required for regulating the data transfer rate to stay within the capabilities of the channel and control unit. These pacing procedures are not unlike those used in data communications. However, they can be simpler because of the simpler nature of the channel's interconnection network. We describe herein a method whereby the channel and control unit exchange information about their capabilities and then dynamically regulate the data transfer rate during each I/O operation.

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

Method And Apparatus for Data Flow Control On a Serial I/O Channel

      In an I/O architecture based on transmission of data in
packets or frames without byte-by-byte handshaking, a flow control or
pacing procedure is required for regulating the data transfer rate to
stay within the capabilities of the channel and control unit. These
pacing procedures are not unlike those used in data communications.
However, they can be simpler because of the simpler nature of the
channel's interconnection network. We describe herein a method
whereby the channel and control unit exchange information about their
capabilities and then dynamically regulate the data transfer rate
during each I/O operation.

      Our method is particularly applicable to the I/O channels of
computers, such as the ESCON* channels of the IBM System/390*
processor family. This is a circuit switching environment in which a
given channel link may be connected to different I/O control units,
with different data rate characteristics, at different times.  It is
also applicable where the interconnections are static but the data
rate characteristics of the channel and/or control unit may change
with time as would be true of a bandwidth-allocating I/O channel.

      There are two components to the method:
   l. Exchange of certain information, called pacing parameters,
between the channel and control unit, which describes their basic
capabilities.
   2. Data request pacing, by which the receiver of data dynamically
regulates the rate at which the sender sends data frames.
Pacing Parameters

      The pacing parameters include information about the data
receiver's instantaneous receiving rate capability, expressed as the
minimum allowed spacing between successive data frames of each
permitted size.  They also include information about the data
sender's capability for accepting requests for continuation of data
transmission during the Data Request Pacing procedure, expressed as
the maximum number of outstanding data requests which the data
receiver may send to the data sender.

      The pacing parameters may be exchanged between the channel and
control unit during system initialization or during each I/O
operation.  There are several design choices, depending on how
dynamically the information in the pacing parameters must be varied.
These possibilities are:
   1. The channel sends its pacing parameters to the control unit
during initialization in the message which requests a logical path,
called Establish Logical Path (ELP).  The control unit sends its
pacing parameters to the channel in the message which accepts each
I/O operation, called Command Response.
   2. The channel and control unit each supply their parameters to
the other when the logical pat...