SUMMARY: network problems??

From: Bruce Kall (kall@mayo.EDU)
Date: Wed Jan 20 1993 - 03:47:40 CST


My original post:

> From bruce Mon Jan 18 14:30:22 1993
> To: sun-managers@eecs.nwu.edu
> Subject: network problems??
> Content-Length: 1087
> X-Lines: 39
>
> Sunmanagers,
>
> What does it mean when spray drops no packets
> from Host A to Host B, but drops >80% of the
> packets from Host B to Host A.
>
> Host A is a 4/330, Host B is a IPX
>
>
>
> When Host A sprays Host B I get:
>
> sending 1162 packets of lnth 86 to Host B ...
> in 0.5 seconds elapsed time,
> no packets dropped by Host B
> 2146 packets/sec, 180.3K bytes/sec
>
> But when I spray from Host B to A I get:
>
> sending 1162 packets of lnth 86 to Host A ...
> in 10.4 seconds elapsed time,
> 955 packets (82.19%) dropped by Host A
> Sent: 111 packets/sec, 9.4K bytes/sec
> Rcvd: 19 packets/sec, 1.7K bytes/sec
>
> One direction is obtaining 180.3K bytes/sec. The
> other way seems to have a problem. Any suggestions?
> Is this normal (an IPX is supposedly faster(?) than
> the 4/330).
>

>From the responses I received, this is normal.

A sampling of the replies:

From: archer!artel.com!myk@uunet.UU.NET

I don't know if an IPX is faster than a 4/330, but this is exactly
what we saw here spraying between 3/50s and SPARCs (which are faster).
Basically, it just means the IPX can process packets faster than the 4/330.

Spraying from a 3/50 to a SPARC was always OK, but when spraying from
a SPARC to a 3/50 the 3/50 always dropped packets, very similar to
what you are seeing.

If you use 'spray -d 1 host' to put a delay between packets, you
should get all the packets.

From: babb@k2.sanders.lockheed.com

An IPX is faster than a 4/330 (~28 MIPS -vs- ~16 MIPS).
What you are seeing can probably be attributed to the IPX being
fast enough to service the spray packets without dropping any,
while the 4/330 can't quite keep up. Moment-to-moment differences
in network traffic could also cause this.

From: stern@sunne.East.Sun.COM

generally, assymetric A->B and B->A spray rates
indicate that A and B send/absorb packets at vastly
different rates, or that there's some piece of
network gear that melts down when stressed in one
direction.

in this case, it means that the IPX is a good bit
faster than the 330. it's sending packets faster
than the 4/330 can take them.

Thanks also to:

ems@ccrl.nj.nec.com
sparc!shevett@erika.bighouse.unipress.com
tommy@boole.att.com
beyea@ERC.MsState.Edu
todd@flex.eng.mcmaster.ca
venus!hammond@sunloop.Central.Sun.COM
ups!upstage!glenn@fourx.Aus.Sun.COM

\------------------------------------------------------------\
 \ Bruce Kall \ Internet: Kall@Mayo.edu \
  \ Mayo Clinic (Foundation) \ Phone: (507)-255-4768 \
   \ Rochester, MN 55905 \ \
    \ \ \
     \ Go BILLS!!!!!!!!!!! \ Go BILLS!!!!!!!!!!!! \
      \------------------------------------------------------------\



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:24 CDT