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

Conference npss::gigaswitch

Title:	GIGAswitch
Notice:	GIGAswitch/FDDI Jan 97 BL3.1 914.0 documentation 412.1ion 412.1
Moderator:	NPSS::MDLYONS

Created:	Wed Jul 29 1992
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	995
Total number of notes:	4519

955.0. "DelayExceededDiscards - How much buffer ?" by COMICS::BUTT (Give me the facts real straight.) Wed Mar 26 1997 05:12

Consider two G/S/FDDI V3.1 each connected to a DAS Conc as follows:-
	 ________	_________	_________	_________
st1-----I	|	|	I	I	I	I	I----stat 2
st3-----I	|	|	IA-----BI	I	I	I----stat 4
st5-----I G/S	|S-----M| CONC	I	I CONC	IM-----SI  G/S	I----stat 6
st7-----I	|	|	IB-----AI	I	I	I----stat 8
st9-----I	|	|	I	I	I	I	I----stat 10
	 ________	_________	_________	_________

The applied load will sometimes saturate the DAS connection. The G/S 's are 
clocking delay exceeded discards on the S ports to the concs.

Now each G/S will see the token at least every TTRT (8ms) and access will be 
shared.

I would expect packets to be discarded due to lack of transmit buffer rather 
than packet aging. 

What is the age timer (2 seconds ?) . Is there really enough transmit buffer 
to allow packets to wait 2 seconds in a queue before they can get on the wire 
when there are so few nodes on the ring ?

Simplistically, lets assume each G/S gets to use 40 Mb/s, if a packet ages 
after 2 seconds it would take 10 MBytes of buffer. How much is allocated per 
port ? 

R.

T.R	Title	User	Personal Name	Date	Lines
955.1		NPSS::RLEBLANC		`Wed Mar 26 1997 08:47`	7
	I'd be interested in seeing the Raw Counters via the DVT monitor for the FGL ports in question. Run test 117 and post the results here or send me mail. Rene'
955.2	Some conters	CHEFS::16.182.96.55::butt	UK CSC Networks	`Tue Apr 01 1997 03:56`	18
	Well.. I tried to get the customer to gather this info. All I have is a fax from one Gigaswitch. The interesting counters are as follows:- in_timed_out 1 out_timed_out 0 in_queue_ovfl cd out_queue_ovfl 0 in_crc_err 1 out_xmit_failed b in-e-bit-err 24 out-cutthru_abort b All others are zero or normal traffic numbers. These numbers don't seem to tie in with the hundreds of "Delay exceeded discards" seen from MCM. I will get someone to recheck this data ! R.
955.3		NPSS::MDLYONS	Michael D. Lyons DTN 226-6943	`Tue Apr 01 1997 09:09`	4
	...well, cd = 205 in decimal - what is MCM reporting for the delay exceeded discard value? MDL
955.4	Somewhat bigger !	CHEFS::16.182.96.55::butt	UK CSC Networks	`Wed Apr 02 1997 05:29`	5
	Ref .3 Frames aged out from MCM = 388680 on the line card = 205 R.
955.5	"dot1dBaseportdelayexceedediscards"	CHEFS::16.182.96.55::butt	UK CSC Networks	`Mon Apr 07 1997 14:04`	15
	I have concerns that the "dot1dBaseportdelayexceedediscards" Mib variable on the Gigaswitch is wrong. The examples I have are as follows:- G/S 1 - in_timed_out = 00 , out_timed_out = 4508 , D_E_Discards = 370523 G/S 2 - in_timed_out = 01 , out_timed_out = 0 , D_E_Discards = 388680 I have taken an SNMP trace and this shows that the value displayed by MCM for "dot1dBaseportdelayexceedediscards" is correct as returned by G/S. Where does the SCP get its "dot1dBaseportdelayexceedediscards" values from? R.
955.6		NPSS::MDLYONS	Michael D. Lyons DTN 226-6943	`Mon Apr 07 1997 16:02`	4
	Please escalate this as an official problem - At the moment I'm working on other problems which have been escalated.... MDL