Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Bandwidth-Based Adaptive Resequencing for Multi-Line Transmission Paths

IP.com Disclosure Number: IPCOM000112550D
Original Publication Date: 1994-May-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 4 page(s) / 167K

Publishing Venue

IBM

Related People

Long, LD: AUTHOR [+3]

Abstract

Developing a multi-line transmission protocol which is efficient, general-purpose, and scalable presents several problems to the designer. Scalability requires that the protocol work efficiently in small environments and that it make maximum use of additional bandwidth investment in large environments. Multi-line support provides incremental scalability but introduces complexities in handling lines of mixed speeds and reliability. In making use of multiple lines, one must contend with the fact that messages may be received in random sequence. Distinguishing between this random sequence and truly "lost" messages in systems that guarantee in-order delivery to the receiving application requires queuing and resequencing of messages. In a scalable protocol, the queuing and detection must be scalable as well.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 28% of the total text.

Bandwidth-Based Adaptive Resequencing for Multi-Line Transmission
Paths

      Developing a multi-line transmission protocol which is
efficient, general-purpose, and scalable presents several problems to
the designer.  Scalability requires that the protocol work
efficiently in small environments and that it make maximum use of
additional bandwidth investment in large environments.  Multi-line
support provides incremental scalability but introduces complexities
in handling lines of mixed speeds and reliability.  In making use of
multiple lines, one must contend with the fact that messages may be
received in random sequence.  Distinguishing between this random
sequence and truly "lost" messages in systems that guarantee in-order
delivery to the receiving application requires queuing and
resequencing of messages.  In a scalable protocol, the queuing and
detection must be scalable as well.

      A common mechanism for handling of out-of-sequence messages is
to hold all messages until the late message is received.  The queue
of waiting messages is typically allowed to grow in size until a
hard-coded upper limit is reached.  The side-effects of this approach
are such that:

o   If the upper-limit is too high, the extra storage that is tied up
    for retaining queued messages can cause performance problems in
    the system.  In systems which are dedicated to message switching,

    the impact is bad enough as is; however, in general purpose
    systems, the unchecked growth of page-fixed (unpageable) storage
    can affect unrelated applications as well and can ultimately lead
    to a costly system outage.

          Some machines dedicated to message switching avoid the
    system outages by providing an overall system throttle for
    storage.  This type of throttle is not practical for general
    purpose systems whose primary purpose is application processing
    (i.e.  database, transaction processing, e-mail, numeric
    computing, etc).  In such general purpose systems, message
    switching is expected to use a minimum of system resources.
    Allowing the resequence queue to grow such that it used a
    majority of the system resources would definitely be considered
    an error.

o   The transmission medium can appear "hung" until the late message
    is received.  This "hang" is generally not perceptible to the
    systems operator until users start complaining about lack of
    response at their system.  It is generally desirable to avoid
    such customer complaints as it may affect the billing practices
    of the organization providing the transmission service.

      Also, in a variable collection of transmission lines the use of
a "hard-coded" limit for out-of-sequence messages results in
suboptimal and potentially costly use of system storage in systems
where this limit is inappropriate.  For example, in a system with a
predefined que...