				pingd README

The code in this directory performs pings on all nodes configured in the 
hostfile and maintains a Network Status File.  

pingd allows the administrator to configure the following in the hostfile:

	The maximum # of pings to send before calling node Not Reachable
	Delay between retransmissions
	UP Cache State timeout - how long to hold the UP state before polling

configure these in the following syntax:

	PING( Retries, Delay, Cache )

by  default, if nothing is specified,

	PING() is the same as specifying PING( 3, 20, 120 )

which will ping up to 3 times separated by 20 seconds before arriving with a 
nodes State, and it will hold an UP state for 120 seconds before polling this
node again.

pingd yields and interesting result; because nodes respond at different times,
a natural skewing of pings occurs over time.  As the ping responses come back,
we mark the node UP, marking the time to send the next ping. 

One change in the algorithm we made was to apply the State Cache Timeout only
to nodes in the UP state.  If the node is NR, the noc wants to know ASAP when 
the node starts responding, so we continually ping nodes that are marked NR.

