Browse Prior Art Database

Method for Improving Network Availability with Redundant Network Servers

IP.com Disclosure Number: IPCOM000118042D
Original Publication Date: 1996-Aug-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 2 page(s) / 118K

Publishing Venue

IBM

Related People

Hunter, SW: AUTHOR

Abstract

Disclosed is a method for improving the availability of network services by using redundant servers in a manner transparent to the network application. An example is given on how this technique could be applied to LAN Emulation.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 50% of the total text.

Method for Improving Network Availability with Redundant Network
Servers

      Disclosed is a method for improving the availability of network
services by using redundant servers in a manner transparent to the
network application.  An example is given on how this technique could
be applied to LAN Emulation.

      This invention solves the problem of a networking client
dynamically switching from a nonoperational networking server to an
operational load sharing server.  This is of particular importance
when high network availability is desired for services such as "LAN
Emulation" and "Classical IP over ATM".  These services use a
networking server to allow a source client to locate a destination
client, but if  a single server is used, the virtual network isn't
fault-tolerant and may have undesirable network availability.
However, if a backup/load-sharing server is available, a source
client can redirect requests (e.g., ARPs) to the backup when the
source client detects a problem with a faulty server.  While this
technique doesn't necessarily  prevent degradation of networking
services to a noticeable level, it is  possible by the increasing the
number of servers to N+1, where is N is  the number of servers needed
to meet desired service.

      The solution proposed is for networking servers to send regular
periodic messages (e.g., SERVER #n ACTIVE) to all networking clients
using their services.  The rate of the SERVER ACTIVE messages can be
negotiated between a client and server when a control VCC is setup
between them (see example below for LAN Emulation).  When a client
quits receiving a SERVER ACTIVE message from a specific server (i.e.,
a timeout occurs due to a faulty server), the client should no longer
send ARP requests to this server, instead they should go to an
alternate known server.  If the SERVER ACTIVE messages resume from
the originally  faulty server, the client should again incorporate it
into its ARP procedure.

      In addition to improving availability, this invention can also
provide performance benefits.  To do this, all SERVER ACTIVE messages
received by a client from all active servers should contain a measure
of the server's internal utilization.  The client can use this
information to select the least utilized server.  This technique
should allow optimal  load balancing across all active servers and
thus provide the best possible server performance for the given
network configuration.

      An Example - The LAN Emulation service gives access to the MAC
services of legacy LANs by being installed on top of the ATM protocol
stack, thus allowing existing LAN packets to traverse the ATM network
while being totally transparent to the end user.  Another service,
chosen by the Internet Engineering Task Force (IETF), is "Classical
IP and ARP  over ATM'".  This service operates at the network layer
and provides internetworking with legacy LANs by using IP routing.
The background, ...