Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1+some 2/3/84; site dual.UUCP
Path: utzoo!watmath!clyde!burl!mgnetp!ihnp4!dual!fair
From: fair@dual.UUCP (Erik E. Fair)
Newsgroups: net.bugs.uucp
Subject: Discrepancies in UUCP SYSLOG timings
Message-ID: <745@dual.UUCP>
Date: Wed, 8-Aug-84 21:34:08 EDT
Article-I.D.: dual.745
Posted: Wed Aug  8 21:34:08 1984
Date-Received: Sat, 11-Aug-84 00:13:14 EDT
Distribution: net
Organization: Dual Systems, Berkeley, CA
Lines: 52

I have been trying to use the information kept in the UUCP SYSLOG for a
generalized report of UUCP traffic. I total up the number of transfers,
number of bytes in both directions and the total time spent
transferring data with awk, and mail myself a report once a week. All
well and good.

Then I modified the awk script to give me average characters per second
for receive and transmit of data, and a big ugly split appeared. We
received data from 10 to 30 cps slower than we transmitted it to the
same site.  I thought there was something wrong with our system until I
set up ucbvax to keep the same statistics. Guess what? They show the
same split for all the sites that they talk to. When ucbvax sends to
the remote, it sends at around 110 cps. When it receives from the
remote, it receives at 80 or 90 cps.

This tended to imply something strange going on in the timing in UUCP
itself. To show you what I mean, here are four lines from the SYSLOGs
of dual and ucbvax:

==> DUAL-SYSLOG <==
ucbvax!daemon M (8/7-9:18:33) (C,7367,1) (0:8:12)  <- 26578 / 245 secs
ucbvax!daemon M (8/7-9:18:38) (C,7367,2) (0:8:17)  <- 68 / 2 secs
ucbvax!uucp M (8/7-17:33:33) (C,12681,1) (0:38:5)  -> 1689 / 15 secs
ucbvax!uucp M (8/7-17:33:39) (C,12681,2) (0:38:11)  -> 57 / 0 secs

==> UCBVAX-SYSLOG <==
daemon dual (8/7-9:19) (460743571) sent data 26578 bytes 243 secs
daemon dual (8/7-9:19) (460743576) sent data 68 bytes 0 secs
uucp dual (8/7-17:34) (460773283) received data 1689 bytes 17 secs
uucp dual (8/7-17:34) (460773288) received data 57 bytes 2 secs

As you can see, there is a two second difference in the times on either
end. The receiving end says it takes 2 seconds longer than the transmit
end claims. For the folks wondering about configurations, DUAL is a
System V R 1 10Mhz 68000, and UCBVAX is a 4.2 BSD DEC VAX-11/750.

I looked over the timing code (it's in each of the protocol modules;
gio.c for this example) and it seems to be reasonable. The only thing
that caught my attention is that there seems to be a lot more packet
code to wade through for receive than transmit. Is that all this is?
Excessive overhead in receive?

Has anyone out there ever profiled uucico for transmitting and
receiving data at 1200 baud?

	discussion of the problem is encouraged,

	Erik E. Fair	ucbvax!fair	fair@ucb-arpa.ARPA

	dual!fair@BERKELEY.ARPA
	{ihnp4,ucbvax,hplabs,decwrl,cbosgd,sun,nsc,apple,pyramid}!dual!fair
	Dual Systems Corporation, Berkeley, California