[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference azur::mcc

Title:DECmcc user notes file. Does not replace IPMT.
Notice:Use IPMT for problems. Newsletter location in note 6187
Moderator:TAEC::BEROUD
Created:Mon Aug 21 1989
Last Modified:Wed Jun 04 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:6497
Total number of notes:27359

4437.0. "NODE4 line statistic, Ethernet Contention, ???" by GWTEAL::WOESTEMEYER (Why??...Why not!!!) Mon Jan 25 1993 12:15

A search of the PA and Node4 documentation produced not explaination of
what the statisitic 'Ethernet Contention' is for a node4 line.  BMS V1.2.


MCC> sh node4 keroon line sva-0 all statis

Node4 keroon Line sva-0
AT 25-JAN-1993 10:04:19 Statistics

Examination of Attributes shows:
                               Duration = 59 Seconds
                   Outbound Utilization = 0 Percent
                    Inbound Utilization = 0.25 Percent
            Average Outbound Block Size = 65.5 Bytes/Block
             Average Inbound Block Size = 96.86 Bytes/Block
                    Outbound Block Rate = 7.000000E-02 Blocks/Sec
                     Inbound Block Rate = 25.39 Blocks/Sec
                       Deferred Percent = 0
               Single Collision Percent = 0
             Multiple Collision Percent = 0
              Inbound Multicast Percent = 100
                          Error Percent = 0
           Count of Data Overrun Blocks = 0
    Count of System Buffers Unavailable = 0
      Count of User Buffers Unavailable = 0
     Count of Collision Detect Failures = 0
Count of Unrecognized Frame Destination = 0
           Count of Frame Send Failures = 0
        Count of Frame Receive Failures = 0
                    Ethernet Contention = 0 Percent        <<<<<<<<<
                         Outbound Bytes = 262 Bytes
                          Inbound Bytes = 145092 Bytes
                   Outbound Data Blocks = 4 Blocks
                    Inbound Data Blocks = 1498 Blocks


Any Takers?

Thanks,
Steve Woestemeyer
CSC/CS - Network Support Group
T.RTitleUserPersonal
Name		DateLines
4437.1Memories of NMCC/DMBIKINI::KRAUSEEuropean NewProductEngineer for MCCTue Jan 26 1993 10:5731
>A search of the PA and Node4 documentation produced not explaination of
>what the statisitic 'Ethernet Contention' is for a node4 line.  BMS V1.2.

This one obviously belongs to the formula chapter of the PA doc,
but isn't there (any doc writers listening?). 

'Ethernet Contention' sounds familiar to me ... and the voice
spoke: "N M C C / D E C n e t M o n i t o r" ;-) 

	Ethernet_Contention = ( bsid + bssc + bsmc ) / dbs

		bsid = Blocks Sent Initially Deferred
		bssc = Blocks Sent Single Collision
		bsmc = Blocks Sent Multiple Collision
		dbs  = Data Blocks Sent

Since the statistics provided by PA are very similar to what
NMCC/DM gave us, I assume that they 'recycled' the formulas (and
made the same mistakes...). PA says Percent so most likely it's 
* 100.

'Ethernet Contention' is a relational value with no unit that
gives an estimate of how busy the Ethernet is, in the view of
this line. A high value means that the line had to wait/retry a
lot before it could successfully send a block. This could be due
to heavy traffic or high collision rates. A low value means that
the line most of the time saw a free Ethernet when it wanted to
transmit. 

*Robert		(who guessed quite a few performance figures
		from NMCC/DM reports :-)
4437.2QAR#465 - MCC013_INTMOLAR::CHRISB::BRIENENNetwork Management Applications!Wed Jan 27 1993 17:3011
Hi Steve,

 Thanks for finding this "oversight" in the Use Book (it DID make it into
the PA MRM, but not many people read those!).

 I've filed a low priority QAR against PA.

						Chris

P.S. Thanks Richard for the description
     in .1 !
4437.3Math ReuseBLUMON::SYLORArchitect = Buzzword GeneratorMon Feb 01 1993 22:3818
    Ah yes, old formulas never die ...
    
    I don't know for sure what PA does, but I'll bet you've got the formula
    (and its lineage) right.
    
    For what it's worth, I theorized that the ethernet contention is
    about equal to the utilization (overall actual utilization of the
    ether). If a line is sending packets at random, and in sufficient
    volume to form a valid statistical sample, then the probablitity the
    packet can't be sent immediately equals the utilization of the cable.
    Little's Law of queueing theory. The packets that don't get sent
    immediately are either "initially deferred", or suffer a collision.
    Hence the formula.
    
    Unfortunately, I never got around to testing that theory against
    practice.... If anyone ever tested the idea, I'd like to hear.
    
    Mark