Martin 53N 1W <ml_news@ddnospamddml1dd.co.uk.dd> writes:

Johan Kullstam wrote:
[...]

cases.  I think that calling it "trial" would mislead people into

[...]

This is where there is somewhat a difference of opinion.

 From a shallow casual quick glance at the last few days, with no
reading of the news, then it appears that the s@h2 project isn't
working. Call that whatever you like as pre-alpha/alpha/beta/non-start
or whatever.

The reality is that the Boinc server and client software has gone
through what I understand as 'alpha' and 'beta' testing in the open
context of this development. Indeed, Boinc is working well for one or
two other projects.


For s@h, as part of the beta test ramp up and getting hammered by the
masses, Berkeley quickly found that their *HARDWARE* utilisation
didn't scale up as anticipated as users joined and quickly became *IO*
bound for the database. In layman's terms, their servers had big fast
brains but their filing system was too slow and cumbersome. The system
then became overwhelmed with the 10000's of users all joining within
hours.


Since then, they've fought through various server software fixes and
workarounds to try to streamline the database loading. They've also
had the hardware break on them a few times, likely because they've
pushed up to limits not normally exercised. They have swapped around
most of their server hardware more than once. All good computer
science and part of the expected development 'gotchas'.

The recent string of problems from live testing have likely pushed
development forwards at a few times the normal pace of trying to debug
offline.

Thank you for your reasonable explaination and take on the set boinc
project.

What has been amazing has been some of the 'users' whinging.

Berkeley are indeed notorious for their poor PR and human
communication about the project. But then again, they are Scientists
and not PR droids. And also, they are American after all (:-P)

I do appreciate them posting on status.  That is what has kept me
going with seti/boinc.

Regardless, s@h is a very good project for the hunt for seti and
consequential science. Berkeley have now added a further important
distributed computing aspect with Boinc opening DC to other projects
more easily.

Yes.  I signed up with the climateprediction folks as per berkeley's
suggestion.  That project also seems worthy.

Funny, but once I added the climateprediction project, the
boinc-client starting downloading from berkeley.  I had been drawing
an endless stream of this:

2004-08-30 22:23:24 [---] Can't resolve hostname  (valid name, no data record of
 requested type)
2004-08-30 22:23:24 [---] Can't resolve hostname  (valid name, no data record of
 requested type)
2004-08-30 22:23:24 [SETI@home] Couldn't start download for /2fd/25au03aa.27140.
29345.953424.199: error -113
2004-08-30 22:23:24 [SETI@home] Couldn't start download for /2fd/25au03aa.27140.
29345.953424.199: error -113
2004-08-30 22:23:24 [SETI@home] Backing off 1 hours, 6 minutes, and 23 seconds o
n transfer of file 25au03aa.27140.29345.953424.199

I am not sure what error -113 is.  It is hard to figure out what a
"bogus work unit" is and how a person should know if they had one.

It is frustrating to install the client and it doesn't seem to be
working and it is difficult to see if the problem is one their end or
my own end.  I did "reset project" a couple of times, that didn't
appear to change anything so I just let it keep going figuring it
would be a problem on their side.

Meanwhile, the faint hearted can stay with s@h classic. Or more
intelligently, you can share your spare CPU cycles with other DC
projects until s@h2 can oncemore usefully soak up the idleness.

I have a cronjob which looks for boinc and if it sees no running boinc
subprocesses, launchs a setiathome classic.  When it does see boinc
action, I touch stop_after_send.txt to let classic finish its unit and
then stop.

Are you in for the Science, the stats, or just for a good whinge?

Oh, all three!

(And I would call this part of the Beta phase of system shakedown
(:-P)

Fair enough.  You already have my opinion.

Happy crunchin',
Martin

-- 
----------   OS? What's that?!
- Martin -   To most people, "Operating System" is unknown & strange.
- 53N 1W -   Mandrake 10.0.1 GNU Linux
----------   http://www.mandrakelinux.com/en-gb/concept.php3

Debian/sid GNU/Linux here.

-- 
Johan KULLSTAM