HP OpenVMS Systems

DECnet NCP Collision Detection counters?

» close window

The Question is:

 
In NCP, there are many counters for a known LINE.  Can you explain what the
following ones mean, and specify a "normal" threshold.  We have 160
PC/terminal users accessing an Alpha 4100 3 CPU 5/400 system with UCX 4.1
and DECNet IV.
 
* Blocks sent, multiple collision
 
* Blocks sent, single collision
 
* Blocks sent, initially deferred
 
* Collision detect check failure
 
After 11600 seconds (3.25 hours) of up-time, they are at the following
levels :
 
multiple collisions - 241
single - 326
deferred - 3836
check failure >65534
 
These seem very high, but are they?
 
Many thanks,
 
Paul Lowe.
 
 

The Answer is :

 
  "Normal" is rather site-specific -- is the current network behaviour
  and performance acceptable?  If so, your current values are "normal".
 
  The counters shown are largely dependant on the LAN traffic, save for
  the last:
 
     'Collision detect check failure'
 
  The collision detection and deferral counts would tend to indicate
  that the LAN segment is very busy, or possibly that it is miswired
  or otherwise misconfigured.
 
  Your local network support organization should be evaluating the
  current configuration and hardware for your network.
 
  The collision detect indicates that the repeater, switch or tranceiver
  that this OpenVMS system is connected is not providing the 'collision
  detect check' signal.  This is not necessarily a particular problem,
  but it does indicate that the network configuration is not entirely
  correct.
 
  There is a possibility that the collisions and the collision detection
  are related, and the collision detect check failure is causing the
  collision counters to increase.
 
  First, determine why the collision detection signal is missing and
  resolve it, then check to see if the LAN activity is actually as heavy
  as it appears to be here.
 

  

     

     
answer written or last revised on ( 29-APR-1999 )
     
» close window