Browse Prior Art Database

Optimized Heartbeat for a Disturbed Cluster Manager

IP.com Disclosure Number: IPCOM000106459D
Original Publication Date: 1993-Nov-01
Included in the Prior Art Database: 2005-Mar-21
Document File: 4 page(s) / 140K

Publishing Venue

IBM

Related People

Azagury, A: AUTHOR [+5]

Abstract

Disclosed is a more efficient heartbeat mechanism with 0(n) signals per heartbeat check, for clusters with transitive and symmetric connectivity.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 100% of the total text.

Optimized Heartbeat for a Disturbed Cluster Manager

      Disclosed is a more efficient heartbeat mechanism with 0(n)
signals per heartbeat check, for clusters with transitive and
symmetric connectivity.

      When several systems cooperate as a cluster, a mechanism for
checking system connectivity has to be provided.  A common mechanism
is the heartbeat; systems exchange periodic signals, and if one of
the systems does not respond to these signals it is considered
disconnected.  There are two basic types of algorithms for heartbeat
check.  The first is for a cluster and is managed by a single cluster
master system that generates all the heartbeats.  If the cluster
master fails to communicate with any system in the cluster, it is
assumed that no other system in the cluster can communicate with that
system.  Each periodic heartbeat check requires 0(n) signals to be
sent in a cluster of n systems.  The second type of algorithm is for
a cluster managed distributively by all the systems in the cluster.
A connection/disconnection between two systems in the cluster does
not necessarily imply their connection/disconnection with the rest of
the systems in the cluster.  Therefore, each system has to send
heartbeat signals to all other systems in the cluster and verify the
results with each of them.