:: 01-Nov-2004 10:24 GMT (Monday) ::

distributed.net is proud to announce the completion of OGR-24!

Four years ago, distributed.net users undertook the search for the optimal
24 mark Golomb Ruler. This year sees the successful conclusion of that
effort. We have proven conclusively by the exhaustive search of all
possible rulers that the currently best known ruler is indeed the Optimal
one.

More precisely it is:
24/9-24-4-1-59-25-7-11-2-10-39-14-3-44-26-8-40-6-21-15-16-19-22

This shortest ruler was found by two independent computers. The initial
report was received on May 24th, 2004 and a second, matching result was
returned on July 3rd, 2004. However it was not until the final stub was
returned and verified that could we rule out the possibility of a
still-shorter ruler. This final stub was returned October 13th, 2004
drawing to a close the complete search of all possible stubs. Due to the
nature of an exhaustive search, distributed.net users have also proven that
the above solution is unique (the ruler’s mirror notwithstanding).

This project was first announced in 1998, started in February 2000, and is
now concluded in 2004. Although 4 years may seem like a long time, the
search was no trivial task. No fewer than 555,529,785,505,835,800 rulers
were checked during that time. Moreover, a second pass of all rulers was
done to rule out (heh) any errors. Additionally a small oversight in the
beginning of the project caused several rulers to be excluded from the
initial search. These were the subject of the much discussed Phase 2
(rulers with initial marks > 70). Incidentally the optimal ruler was
amongst these. (“9+24+4+1+59 > 70”) The double phase, phase 2 and their
verification each required additional structural changes which also
contributed to the overall 4 year duration.

Note that distributed.net users continue to pursue the solution to the
OGR-25 project which began in parallel with OGR-24. We have currently
completed 10-15% of OGR-25 phase 2 which is about 65% overall.

To celebrate the successful end of yet another distributed.net project all
our contributors are invited for a drink…when we find a place large
enough to host the 41,805 people that participated in this particular
distributed effort. :)

The shortest ruler was first found by Matt Richards (Matt_R in
#distributed). It was then confirmed by Mitsuru Aoki of the SEGA Users
Group Team (#1958). The final stub was returned by Sebastian “Pax”
Schmitz. We’ll be sending them some free distributed.net swag and shirts
for their noteworthy contributions to the project.

Related Links:
– http://www.distributed.net/ogr/
– http://n0cgi.distributed.net/statistics/ogr/ogr24-all-nodes-day.png
– http://n0cgi.distributed.net/statistics/ogr/ogr24p2-percent.png

Thank you all and keep those computers busy!

:: 17-Jul-2004 09:47 GMT (Saturday) ::

The fine folks at hackerthreads.com are now selling some really cool t-shirts
with the distributed.net cow on them. I’ve updated the dnetware web page to
point to their distributed.net affiliate storefront, but you can go straight
there from here:

http://slacker.com/newshirts.html

Sport some stylish fashion and help out d.net at the same time.

(the “Moo Country” shirt is my favorite, don’t miss the image on the back)

:: 22-May-2004 15:38 GMT (Saturday) ::

iGive seems to be back with a frenzy. If you’ll recall, distributed.net
was able to raise a lot of money through iGive shopping in the earlier
years of the organization. It looks like iGive is back after a period of
inactivity and they’re offering their $5 per new referral bonus again.

If you want a free way to help us recover from the statsbox purchase, click
over to http://www.iGive.com/DistributedComputing and sign up (or refresh
your old membership). Basically how it works is that if you are an iGive
member and you make an online purchase with one of the partner merchants,
distributed.net gets a tax-free percentage of the purchase price at no cost
to you.

Merchants include Dell, Barnes and Noble, and OfficeMax.

In any event, enjoy the weekend and get your clients upgraded to the OGRp2
builds!

:: 11-Apr-2004 14:51 GMT (Sunday) ::

The new statsbox is built out and I’ve got the backup of the statsdb loaded
and audited. Apache and PHP are up and running and I’ve got our admin
scripts loaded and going.

I’ve done some preliminary tuning of PostgreSQL and I’m not seeing any
disk-based sort_memory used during the normal stats processing process.

Some timing comparisons between blower and the new box:

Sample run from blower’s logs (45 minutes total):
00:49 (statsbox-iii/r72) Beginning daily processing routines
01:10 (statsbox-iii/r72) Daily processing for 20040121 has completed
01:10 (statsbox-iii/ogr) Beginning daily processing routines
01:34 (statsbox-iii/ogr) Daily processing for 20040121 has completed

On the new box (9 minutes total):
19:40 (statsbox-iv/r72) Beginning daily processing routines
19:46 (statsbox-iv/r72) Daily processing for 20040205 has completed
19:46 (statsbox-iv/ogr) Beginning daily processing routines
19:49 (statsbox-iv/ogr) Daily processing for 20040205 has completed

I’m seeing raw hourly log import times in the 20-30 second range. Overall
it’s taking the new box ~32 minutes to import an entire day’s worth of logs
for both projects combined.

This is giving us a total daily processing time of 41 minutes with log
saturation from the keymaster. At that rate, stats should be current
before too long.

Thanks again for your patience, everyone. We’re plugging away at getting
this box up and available again.

:: 07-Apr-2004 16:47 GMT (Wednesday) ::

The new statsbox has arrived from asacomputers.com — It’s plugged in and I’ve
got FreeBSD installed and running (albeit a bit bare at the moment). We’re
just going to chug through getting everything running and our data loaded into
the database over the next day or so. Assuming everything goes smoothly, we’ll
have stats back up in short order.

Front view, times two, open and closed:
http://www.slacker.com/photos/computers/open_and_closed

Top down, deconstructed:
http://www.slacker.com/photos/computers/inside

It’s a sweet little box (but loud!)

Mem: 15M Active, 375M Inact, 150M Wired, 32K Cache, 199M Buf, 3231M Free
Swap: 8192M Total, 8192M Free

Filesystem Size Used Avail Capacity Mounted on
/dev/twed0s1a 989M 25M 885M 3% /
devfs 1.0K 1.0K 0B 100% /dev
/dev/twed0s1g 40G 1.4M 37G 0% /home
/dev/twed0s1e 3.9G 12K 3.6G 0% /tmp
/dev/twed0s1f 97G 893M 88G 1% /usr
/dev/twed1s1d 541G 4.0K 498G 0% /usr/local/raid10
/dev/twed0s1d 31G 862K 29G 0% /var

CPU: AMD Opteron(tm) Processor 244 (1793.33-MHz K8-class CPU)
Origin = “AuthenticAMD” Id = 0xf58 Stepping = 8
Features=0x78bfbff
AMD Features=0xe0500000
real memory = 4211015680 (4015 MB)
avail memory = 4006965248 (3821 MB)
FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs

:: 02-Apr-2004 09:49 GMT (Friday) ::

Last night BovineOne brought the old, dead statsbox over to my house to see
what I could do with it. Moose has already exhausted all the recovery
options I’d know about (and quite a few more, I’m sure) with no luck.

I thought maybe I could at least get Knoppix booted up so I could get dnetc
going again, but had little luck. Knoppix panics trying to probe the SCSI
adapters.

At the very least I’ll be able to pull the database dump off the DLT,
although I think that it’s older than the copy I’m working with currently
from Decibel.

http://www.slacker.com/photos/computers/IMG_1792 for a pic of Slacker NOC,
if anyone’s curious.

As far as I know, we’re still looking at a Tuesday ship date on the new
server. I’ve asked if that date could be pushed up, but no word yet from
the vendor.

Lastly, I got bored yesterday and coded up an RSS feed of .plans. which is
available at http://n0cgi.distributed.net/cgi/rss-plans.cgi or on
LiveJournal at http://www.livejournal.com/userinfo.bml?user=dnetplans

Moo.

:: 01-Apr-2004 09:51 GMT (Thursday) ::

Since the delivery timeline on the Dual Opteron box we ordered keeps
slipping and we realize that it’s crucially important that we gets stats
back up and running in a reasonable timeframe, I’ve just cancelled the
order on the Opteron box. We all decided it would make more sense to buy a
machine locally so we wouldn’t have to wait for shipping delays.

This morning I headed over to CompUSA and picked up a nice eMachines
Minitower. Since we’d lose the 90 days of telephone support if we
installed FreeBSD on the box, we’re just going to stick with Windows ME
which was preinstalled on the machine.

As soon as I’ve had a chance to get Access installed we should be ready to
start serving up stats again. Thanks again, everyone, for your patience
through this unexpected downtime. I’m excited about getting something up
and running now instead of just waiting for that other box!

(By the way, does anyone know the proper modem initialization string for
the eMachines winmodem? I need to get the box online asap)

:: 25-Mar-2004 23:06 GMT (Thursday) ::

Just to answer some questions… Apologies for the acronym overload…

The old statsbox (aka “blower” or “statsbox3”) was a quad xeon 450mhz
with 2GB RAM and Dell Perc RAID. It had five SCSI disks configured in a
2xRAID1 3xRAID5 configuration.

The new statsbox has been ordered (as yet unnamed). It will be:

Dual Opteron 1.8GHz
4GB RAM
8X200GB SATA (Hotswap) on 3Ware 8506 RAID Controller
3U Case
No keyboard :)

The current plan is to split the 8 drives as:
2xRAID1 + Hot Spare
4xRAID10 + Hot Spare

We’ll of course be staying with PostgreSQL and FreeBSD, although bumping
to FreeBSD 5.x for amd64 support.

I expect to get an ETA for delivery tomorrow (Friday 26-March UTC-6)