Browse Prior Art Database

Algorithmic Method for Distributing Clients Predictably and Evenly to Servers

IP.com Disclosure Number: IPCOM000103599D
Original Publication Date: 1993-Jan-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 2 page(s) / 74K

Publishing Venue

IBM

Related People

Stringfellow, DB: AUTHOR

Abstract

This article describes a methodology in which clients are assigned to servers with an even distribution where clients and servers mutually determine who services who with no negotiation communication overhead.

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

Algorithmic Method for Distributing Clients Predictably and Evenly to Servers

       This article describes a methodology in which clients are
assigned to servers with an even distribution where clients and
servers mutually determine who services who with no negotiation
communication overhead.

      When a group of clients have the regular need to request work
from a group of servers, two problems generally emerge with regard to
which server a client should communicate with.  A simple, yet widely
used, approach to determine which server a client will use is to send
a message to a known or dedicated server, requesting where future
communications should be directed.  The problem with this method is
that bottlenecks through this known server are highly probable and a
network can be overwhelmed with request and direction messages.  A
second problem that exists in an environment where many clients are
communicating with many servers, is the even distribution of clients
to servers.  Both of these problems are addressed with this
invention.

      The invention and solution to the above mentioned problem is
found in an algorithm which independently derives an equal
distribution of clients to servers.  This is to say, the client is
able to determine which server it should use through an algorithm,
thereby eliminating direction communications.  This algorithmic
approach has been tested with a variety of data names and has been
found to always distribute the data names to a number of servers in
an equal fashion.

      Essentially, this algorithm splits a character string into
multiple smaller character strings.  These smaller strings are
interpreted as binary numbers and these numbers are summed together.
This sum is...