The December CACTUS meeting will be held at 7:00pm (6:30pm for pizza and lively, informal discussion) on Thursday, December 18th, 2003 in the auditorium of UT Applied Research Laboratories (see below for directions to the facility).
Please remember that in January we will be holding elections for CACTUS officer positions.
by Lindsay Haisley
It's interesting, and at this point heartening, to see the whole SCO/Linux/Unix business playing out like a Shakespearean comedy, almost predictably, it seems. I've been paying daily visits to Pamela Jones' excellent Groklaw website (http://www.groklaw.net/) which has lead the way, in a truly Open Source fashion, in shining the clear light of day on all of SCO's FUD and their legal and PR thrashing about. Much of what shows up, even in the general technical and business press about the issue turns out to be less than complete and lacking in depth compared to PJ's website which is maintained with the help of a crew of hard-working volunteers with legal experience.
For those of you who haven't been following the slowly-unfolding drama, SCO suffered a legal setback last week in their attempts to say "you go first" to IBM in laying potentially infringing code on the table. The judge in the case indicated that if SCO claims IP infringement, the burden is clearly on them to set forth exactly what it is that they claim is infringing, and gave them 30 days to put their cards on the table. Finally, it seems, we will get to see the mysterious and elusive infringing code about which SCO has given us so many press releases.
Darl McBride continues with his monthly "fireside chat" press releases berating the GNU General Public License as "unconstitutional" and a threat to free enterprise, but all of this is beginning to look a bit thin to even the mainstream trade press. The Royal Bank of Canada, which, along with BayStar Capital, invested $50M in SCO recently, has begun to feel the chill, starting at their feet, and is making subtle moves to insure that their investment doesn't end up in the hands of David Boies and friends, who have some pretty heavy hooks into SCO's equity should the tide turn against them. McBride no longer seems to be able to shore up SCO's sagging stock price with fresh rounds of bluff and bluster.
So there is encouraging news, it seems. With hard work and cooperation, the truth will win out!
by Lindsay Haisley
First, I ran a very simple perl program designed to exercise the CPU and see just how fast each machine can shuffle bits around in memory.
$ cat perl-timer.pl
#!/usr/bin/perl
for ($i=1; $i<=1000000; $i++) {
}
All this does is count to 1 million and quit. Here are the results of
running this on the linux.cactus.org and sun1.cactus.org.
On the Linux box:
$ time ./perl-timer.pl real 0m1.407s user 0m1.410s sys 0m0.000s
Not too bad! Here's the result of running the same program on sun1:
$ time ./perl-timer.pl real 0m2.491s user 0m2.460s sys 0m0.020s
The Linux box is about a third faster than sun1.
So how about I/O issues. The results are very interesting! I wrote a similar perl script to repeatedly write a short file to disk.
#!/usr/bin/perl
$sometext = <<EOF
[an old CACTUS newsletter included here]
EOF
;
for ($i=1; $i<=50; $i++) {
open(FROZ, ">froz");
printf(FROZ "%s", $sometext);
close(FROZ);
}
Running this on linux.cactus.org, we get the following results:
$ time ./perl-timer2.pl real 0m0.043s user 0m0.030s sys 0m0.020s
However on sun1, the program takes much longer!
$ time ./perl-timer2.pl real 0m18.879s user 0m0.030s sys 0m0.040s
As you can see, when I/O is involved, there's a substantial performance difference! But wait! It seems that our home directories on this box are NFS mounted from a Netapp box (backup-800.corenap.com) and we all know about NFS, yes? So instead of doing our I/O to a home directory, we'll write to /tmp/froz rather than ~/froz. Things get a lot faster. We can safely bump the number of read/write iterations to 1000 without having to go take a nap. Here are the results. First the Linux box:
$ time ./perl-timer2.pl real 0m0.498s user 0m0.260s sys 0m0.240s
And sun1 .....
$ time ./perl-timer2.pl real 0m0.703s user 0m0.250s sys 0m0.450s
So our Linux box is still faster, but the new Sun is no slouch. The story has a moral, though. If you want to do any fast data processing on the Sun, do your file I/O to /tmp or some other public directory. You'll be glad you did!
[Editor's note: This has empirically determined something I've felt all along -- in deference to the "Sun bigots" out there, Solaris (or the Sun hardware) has always seemed to be significantly slower than any of the *BSD and Linux operating systems out there. On a 533 MHz P3 running FreeBSD 4.8-RELEASE and a 3ware Escalade dual IDE RAID controller, here are the numbers I got:
ns9$ time ./perl-timer.pl real 0m1.229s user 0m1.080s sys 0m0.008s ns9$ time ./perl-timer2.pl real 0m0.095s user 0m0.023s sys 0m0.023s
So not only is FreeBSD no slouch, but it too gives a Sun system a run for its money. To be "fair," I also ran this on a multiprocessor SunFire-880, and here's what I got:
csdsun1$ time /home/gil/perl-timer.pl real 0m0.689s user 0m0.670s sys 0m0.010s csdsun1$ time /home/gil/perl-timer2.pl real 0m0.359s user 0m0.030s sys 0m0.030s
So, as you can see, if you want real performance on a Sun system, you can definitely get it using their high-end hardware!! However, my 1 GHz P3 laptop running FreeBSD beat out the SunFire on the first timer, and even on a NFS-mounted filesystem, the second test ran with 2.397s of real time on the laptop. These kinds of numbers should be something that Sun engineers take seriously. ...Gil]
by Luis Basto
[Editor's note: We want to thank Gil Kloepfer for renewing his membership last month]
Calm and slightly cloudy with occasional thunderstorms from the west.
Somehow people are using the CACTUS P.O. address as a clearinghouse for bankruptcies. This month we received two notices from Bankruptcy Services to Floormaster addressed to the CACTUS P.O. Box. Perhaps I should re-direct these to SCO.
Membership
To renew your membership, please send check or money order payable to CACTUS ($25/yr for regular membership and $96/yr for corporate sponsorship):
CACTUS
PO BOX 9786
AUSTIN, TX 78766-9786
Applied Research Laboratories/University of Texas at Austin (http://www.arlut.utexas.edu/)
- (Gil Kloepfer, Computer Science Division (CSD), 835-3771, gil [at] arlut <dot> utexas <dot> edu)
CoreNAP, L.P. (http://www.corenap.com/)
- (Kenneth Smith, (512) 685-0010, kenneth [at] corenap <dot> com)
- "Providing Austin and central Texas businesses and power users the best choice for server colocation and high speed Internet access."
OnRamp (http://www.onr.com/)
- Internet service provider.
Outserv.net (http://www.outserv.net/)
- IT operations and management solutions to small and midsized businesses.
Covad/Laserlink (http://www.laserlink.net/)
- (Chip Rosenthal)
Dresser Industries - Wayne Division (http://www.wayne.com/)
- (Steve Cox, steve <dot> cox [at] dresser <dot> com, (512) 338-8444)
A leading supplier of integrated retail solutions to the global petroleum and convenience store industries, including point-of-sale systems, fuel dispensers, and after-sale support services.Journyx (http://www.journyx.com/)
- Provider of workforce management software and services
Prog Corp.
- (David Mallis, prog [at] cactus <dot> org, (512) 451-7191)
- Develops educational materials used for in-service training, classroom teaching, and independent study. They also provides consulting services for instructional program design, development, and implementation.
VoIPing, LLC (http://www.voiping.com/)
- A Central Texas privately owned and operated partnership specializing in IT Consulting and Services. (Email info [at] voiping <dot> com. Phone 512-698-VOIP (8647) or 512-698-8031)
Applied Formal Methods, Inc.
- (Susan Gerhart, 794-9732, gerhart [at] cactus <dot> org)
Auspex Systems (http://www.auspex.com/)
- Fastest reliable network fileservers.
Austin Code Works
- (Scott Guthery, 258-0785, info [at] acw <dot> com)
BestRegistrar.com (http://www.bestregistrar.com/)
- (Steve Locke, (800) 977-3475), swl [at] cas-com <dot> net)
- A top-level domain name registrar, CORE member.
CTG
- (Maurine Mecer, 502-0190 [FAX 502-0287])
- Professional recruiting.
EDP Contract Services
- (Mark Grabenhorst, 346-1040) Professional recruiting.
Hewlett Packard (http://www.hp.com/)
- (Bill Sumrall, 338-7221)
Hounix (http://www.texascomputers.com/hounix/)
- (Marilyn Harper)
- Houston's Unix Users Group.
Network Appliance Corporation (http://www.netapp.com/)
- (Frank Mozina, fmozina [at] netapp <dot> com)
O'Keefe Search (http://www.okeefesearch.com/)
- (John O'Keefe, john [at] okeefesearch <dot> com, 512-658-9224 or 888-446-2137)
- Professional recuiting.
Sailaway System Design
- (Chris J Johnson, 447-5243)
Schlumberger (http://www.slb.com/)
- (Kathy O'Brien, obrien [at] asc <dot> slb <dot> com)
- Technical services and products in over 100 countries.
Silicon Graphics (http://www.sgi.com/)
- (Don Williams, 346-9342)
Solid Systems
- (Pete Farrell, 442-2222)
Sterling Infomation Group (http://www.sterinfo.com/)
- (Darrell Hanshaw, 344-1005, dhanshaw [at] sterinfo <dot> com)
Sun Microsystems (http://www.sun.com/)
- (Rick Taylor)
- Supplier of Unix client-server computing solutions.
Texas Internet Consulting (http://www.tic.com/)
- (Smoot Carl-Mitchell, 451-6176, smoot [at] tic <dot> com)
- TCP/IP networking, Unix, and open systems standards.
Technow
- A Sun Authorized Training Center and a Hardware Reseller.
Unison Software
- (Shelley St. John, 478-0611)
- Supplier of networked systems management solutions.
UT Computer Science Department
- (Patti Spencer)
UT Computation Center
- (Mike Cerda, 471-3241, cerda [at] uts <dot> cc <dot> utexas <dot> edu)
Please do not show up earlier than 6:20 pm on the specified day. Enter through the main entrance at 10000 Burnet Road for ARL:UT. Tell the guard that you are here for the CACTUS meeting. You will be required to sign a log book, but not required to wear a badge. The guards will direct you to the auditorium entrance. Limited parking in the front of the building is available, but more extensive parking is available in the large parking lot just north of the ARL building. After 6:30 pm, all entrances to JJ PRC, except for the Burnet Road entrance, are closed and locked. You can still enter the parking lot in front of the ARL building. No parking tags are necessary after 6:00 pm. See map for further details.
Online maps are available at: