:: 10-Dec-2002 16:55 GMT (Tuesday) ::

The team summary page is now working for RC5-72, thanks to the efforts Chris
Bolt, one of our participants.

There are still bugs to go around, as I mentioned in my previous .plan.
Also, it turns out that most people don’t have access to re-assign bugs to
themselves, so instead of re-assigning it as I mentioned in my .plan, just post
a comment and add yourself to the CC list so that people can see the bug is
being worked on and no one wastes time. :)

:: 08-Dec-2002 23:20 GMT (Sunday) ::

A small amount of progress has been made on RC5-72 stats this weekend. The main
summary page and the participant summary page have been modified to display the
correct data.

http://stats.distributed.net/projects.php?project_id=8
http://stats.distributed.net/participant/psummary.php?project_id=8&id=39622

There are still many pages left to modify, though, and this is where YOU can
help!

I’ve opened a meta-bug that has dependencies on pages that still have to be
modified. You can see this bug at

http://n0cgi.distributed.net/bugs/show_bug.cgi?id=2980

and the dependency tree is at

http://n0cgi.distributed.net/bugs/showdependencytree.cgi?id=2980

All of the bugs that bug 2980 depend on are pages that require the changes
outlined in bug 2980. If you have some PHP knowledge and would like to help,
here’s what you need to do:

1) Re-assign the bug for the page you want to work on to yourself (you’ll need
an account in bugzilla to do this) 2) Make the changes 3) Attach the new page,
or a patch, to the bug and re-assign it back to stats-bugs so that one of us
can commit the fixes

There are about 6 pages that need to be changed that I haven’t opened bugs on
yet; I want to clean them up a bit before setting others loose on them. If you
want to help and all the existing pages have been ‘taken’ by someone already,
just catch me on IRC (DeciBull on irc.distributed.net) and I’ll set you up. IRC
is also the best way to get questions answered, though you can also email me.

Moo!

:: 05-Dec-2002 16:56 GMT (Thursday) ::

Quick update on RC5-72 stats…

First, there was some mis-communication internally about intended launch dates.
This is why we ended up launching before stats were ready. The changes required
to get RC5-72 stats aren’t really major, but we wanted to take this opportunity
to make the stats system easier to maintain.

A year or two ago, we made a major effort to revamp stats so that there
was a single code-base for all the projects. This has helped tremendously
with ongoing maintenance, but there were still a few bits of code that were
hard-coded specifically for RC5-64, namely the code that reports how many
keys were done. Right now it’s not only hard-coded, but it’s also hard-coded
assuming that what’s stored in the database is 2^28 blocks.

Rather than compound the problem by hacking hard-coded support for RC5-72 and
it’s 2^32 blocks into the PHP, we are now in the process of changing things so
that the PHP can find out from the database exactly how to present information
on the web pages, in terms of what units are stored in the database and what
multiplication/division needs to be done to make what’s in the database a bit
more presentable.

We also decided that for constancy, it would be best if we stored the number .
of keys done in the database, instead of blocks, because we store nodes for .
OGR and not stubs The database code has already been modified to reflect this,.
and the data stored in the database for RC5-72 is now in keys. Since the PHP .
code is still expecting 2^28 blocks though, this obviously means you can’t .
trust any numbers the pages are showing .

What remains to be done at this point is to go through all the PHP code and
remove the hard-coded RC5 stuff and replace it with code that pulls the
required information dynamically from the database. Paul is working on that
right now, but I don’t have any ETA. As he commits this code you will start
seeing OGR pages that have both Gnodes and full node counts, to roughly match
keys and blocks in RC5.

Thanks as always for your patience.

:: 29-Oct-2002 11:12 GMT (Tuesday) ::

Stats are back up.

:: 29-Oct-2002 10:37 GMT (Tuesday) ::

Stats will be down for a few minutes while I upgrade blower to FreeBSD 4.7

:: 07-Oct-2002 20:52 GMT (Monday) ::

http://stats.distributed.net/projects.php?project_id=205

}:8)

:: 04-Oct-2002 16:43 GMT (Friday) ::

This hasn’t been a fun week in stats-land, as many of you can probably tell.

I started copying the rc5 data. At first, I did all of 1997 with no problem. I
tried doing 1998 the same way, but the database transaction log filled up, so I
went to copying each day instead. Somewhere around January of 1999 it started
getting rather slow. It was taking nearly 10 minutes to copy a single day by
the time it hit Sept. 9, 1999, so I pulled the plug on that and decided to bcp
the data instead.

Things went downhill from there. First there were some problems with the BCP,
which I didn’t notice until it was loaded into the main table, so I had to
delete everything I just added. That finished last night (as well as finishing
a wayward OGR run). This morning, after grabbing a second backup of the most
critical tables, I started bcping the data that I had checked to make sure it
was valid into the main table again.

A while later (after the BCP had run it’s course), Bruce noticed that what I
had just imported was an exact duplicate of the existing RC5 data… including
the wrong project ID.

So, I’m now in the process of blowing away all the bad data from the main
table. After that’s done, I’ll have to import two copies of it.

The good news is that should be done sometime tomorrow, barring further
screw-ups. The bad news is that there won’t be any OGR updates until this is
taken care of.

Sorry for the inconvenience.

:: 27-Sep-2002 19:20 GMT (Friday) ::

Quick update: turns out doing this data copy with the website turned on is
making Sybase pretty mad, so I’ve turn the web off for now. I’ll try turning it
back on with the p-history stuff removed (which I think is the main cause of
the problem), but that might not work.

This might delay getting the second set of RC5 stats up.

:: 27-Sep-2002 18:59 GMT (Friday) ::

Well, I haven’t read RC5@ yet, but I know that a number of people are very
unhappy with the decision to remove all the stats from after the winning key
was found. This was done because it’s what we’ve always done in the past, and
is arguable the most correct thing to do, since there is definite value in
knowing the exact amount of effort that was required to find the winning key.

On the other hand, we certainly understand why many of you would like to see
stats for all the work that was done, including work after the key was found.
In order to try and make everyone happy, we’re going to post both sets of
stats. One set will stop at the time the key was found, the other set will go
to the end of the logs.

Blower is working on copying the data right now. It’s slowing the box down
quite a bit, so that’s why stats are slow right now. Since I deleted the raw
data from stats, it will have to re-process the logs from after the key was
found. I estimate that effort will take about 10 days, baring any problems.

:: 27-Sep-2002 02:22 GMT (Friday) ::

Stats are back on, and I’m going to run 7/14 now. Everything should be back to
normal in a few hours.