Conference azur::mcc

813.0. "aver. block size" by JETSAM::WOODCOCK () Tue Mar 19 1991 13:10

Hi there,

Below is the output file for a procedure for calculating % utilization
and message sizes for given intervals (this one is for 10:30-10:45). The
message size is a reflection of the actual user data (ie no DECnet header).

The formula used is:	(bytes_in/blocks_in)-54

The subtraction of 54 is to compensate for overhead. 54 is used instead of
64 because I was under the impression that the 10 byte DDCMP header is 
already stripped when looking at the counters (bytes_in). Therefore the actual
packet size averages are 41.2 + 64 = 105.2 (inbound) and
213.4 + 64 = 277.4 (outbound). This makes for about a 10 byte difference 
between the algorithm used here and MCC's.

The MCC formula doc states that utilization= ((bs+hs*dbs) / sp*time)*bits*100
where hs is the header size (in this case 10 bytes). This indicates that
compensation for the header is made for utilization because the bytes_sent 
counter does not reflect the header bytes for each block sent.

But the formula used for average block size is bs/dbs. Should this also
be compensated by hs to have a formula like the following or am I missing
something here????

		aver out block size = bs/dbs + hs

brad...

..............................................................................

                          REAL TIME UTILIZATION RESULTS
                         _______________________________


   Circuit Information        
  _____________________

                  Node: BBPK01	  
        Circuit Device: SYN-0	  
         Circuit Speed: 384000	  
         Adjacent Node: AKO	  


   Time        Inbound     Outbound     Average In     Average Out     Circuit
   Stamp          %            %        Data Size       Data Size       Downs
________________________________________________________________________________
19-MAR 10:45    6.595        19.25        41.20          213.4            0


...............................................................................

show node4 bbpk01 cir syn-0 all statis,for start 10:30 every 00:15 
duration 00:15 until 10:45,in domain .pko-24
!
!Node4 bbpk01 Circuit syn-0 
!AT 19-MAR-1991 10:45:00 Statistics
!
!Examination of Attributes shows:
!                               Duration = 900 Seconds
!                   Outbound Utilization = 19.21 Percent
!                    Inbound Utilization = 6.58 Percent
!            Average Outbound Block Size = 268.35 Bytes
!             Average Inbound Block Size = 94.93 Bytes
!                    Outbound Block Rate = 33.13 Blocks/Sec
!                     Inbound Block Rate = 30.09 Blocks/Sec
!                   Outbound Packet Rate = 33.11 Packets/Sec
!                    Inbound Packet Rate = 30.07 Packets/Sec
!       Originating Packets Sent Percent = 0
!   Terminating Packets Received Percent = 0
!                      Outbound Overhead = 5.000000E-02 Percent
!                       Inbound Overhead = 5.000000E-02 Percent
!       Outbound Congestion Loss Percent = 0
!        Inbound Congestion Loss Percent = -- Invalid Value
!                 Count of Circuits Down = 0
!Count of Circuit Initialization Failures = 0
!          Count of Local Reply Timeouts = 0
!         Count of Remote Reply Timeouts = 0
!  Count of Circuit Outbound Data Errors = 0
!   Count of Circuit Inbound Data Errors = 0
!           Count of Local Buffer Errors = 0
!          Count of Remote Buffer Errors = 0
!            Count of Selection Timeouts = 0
!                          Error Percent = 0
!    Count of Forwarding Congestion Loss = 0
!
!

>>The MCC formula doc states that utilization= ((bs+hs*dbs) / sp*time)*bits*100
>>where hs is the header size (in this case 10 bytes). This indicates that
>>compensation for the header is made for utilization because the bytes_sent 
>>counter does not reflect the header bytes for each block sent.

>>But the formula used for average block size is bs/dbs. Should this also
>>be compensated by hs to have a formula like the following or am I missing
>>something here????

>>		aver out block size = bs/dbs + hs


I see your point. The Phase IV statistics formulae were taken from
NMCC/DECnet Monitor. We will have to check this with the people who
defined them. 

    I believe the statistic, average  outbound  block  size  (bs/dbs),  is
    intended to measure the size of the 'user data block' at the data link
    layer.  'User' here refers to the user of the data link layer.