Raise of minimum required BOINC Core Client version

log in

Advanced search

Message boards : Technical News : Raise of minimum required BOINC Core Client version

1 · 2 · Next
Author Message
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3611
Credit: 128,377,897
RAC: 55,563
Message 107851 - Posted: 16 Nov 2010, 12:44:57 UTC

This is an early pre-warning: At some point in the next weeks or months, but definitely before the end of the upcoming GW run S5GC1HF we will raise the minimum BOINC Core Client version required for Einstein@home from 4.19 to 5.8.0. People using older Core Clients will need to upgrade to get work then.

BM

mikey
Avatar
Send message
Joined: 22 Jan 05
Posts: 2573
Credit: 160,453,930
RAC: 127,246
Message 107852 - Posted: 16 Nov 2010, 12:54:56 UTC - in response to Message 107851.

This is an early pre-warning: At some point in the next weeks or months, but definitely before the end of the upcoming GW run S5GC1HF we will raise the minimum BOINC Core Client version required for Einstein@home from 4.19 to 5.8.0. People using older Core Clients will need to upgrade to get work then.

BM


You might want to put this on the front page and email it to all those using older versions. I know SOME people feel they cannot upgrade due to using their pc's at work, so you might lose some folks over this. Others will take the chance to do better Science! I would guess though that alot of us are already past that point. Good luck and I hope for the best!!
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3611
Credit: 128,377,897
RAC: 55,563
Message 107858 - Posted: 16 Nov 2010, 14:42:05 UTC - in response to Message 107852.
Last modified: 16 Nov 2010, 14:44:25 UTC

You might want to put this on the front page and email it to all those using older versions.


I intended to do this when I have a bit clearer idea of the time we actually will do that.


I know SOME people feel they cannot upgrade due to using their pc's at work, so you might lose some folks over this.


Well, you should run BOINC "only on computers that you own, or for which you have obtained the owner's permission." What I'm more worried about are the people that run the stock client of their Linux distributions, some of which are pretty conservative. This includes some LSC computing clusters, btw.


Others will take the chance to do better Science!


The main feature that we need is the feature to compress the result file before it's sent back to the server. Currently the
BM
.clair.
Send message
Joined: 20 Nov 06
Posts: 50
Credit: 11,162,161
RAC: 0
Message 107870 - Posted: 16 Nov 2010, 22:11:47 UTC

And here was me thinking that my pair of win98se crunchers running 5.10.45 was an `old` version :)
The rest are on 6+.

mikey
Avatar
Send message
Joined: 22 Jan 05
Posts: 2573
Credit: 160,453,930
RAC: 127,246
Message 107883 - Posted: 17 Nov 2010, 11:31:30 UTC - in response to Message 107858.

You might want to put this on the front page and email it to all those using older versions.


I intended to do this when I have a bit clearer idea of the time we actually will do that.


I saw you did it, it looks nice!

I know SOME people feel they cannot upgrade due to using their pc's at work, so you might lose some folks over this.


Well, you should run BOINC "only on computers that you own, or for which you have obtained the owner's permission." What I'm more worried about are the people that run the stock client of their Linux distributions, some of which are pretty conservative. This includes some LSC computing clusters, btw.


It totally agree but you know people!

Others will take the chance to do better Science!


The main feature that we need is the feature to compress the result file before it's sent back to the server. Currently the <10% of result files that are not compressed are taking up more than half of the storage space, I/O bandwith etc. on the server side. (We once used to compress the result files in the application, but BOINC dropped support for the boinc_zip library long ago and for good reason).
BM


Hmmm you learn something every day, I now know what I learned today!! Thanks!!

But why stop at version 5.8? Why not make people go all the way to 6.x instead? As stated by clive alot of us are already into the 6.x versions anyway. Versions 6.10.58 is the recommended version now and 6.12.6 is the latest development version available. I think 6.4.5 is the minimum to be able to use a gpu while crunching. And you are working on a new version of that software.
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3611
Credit: 128,377,897
RAC: 55,563
Message 107884 - Posted: 17 Nov 2010, 11:43:41 UTC - in response to Message 107883.

But why stop at version 5.8?


For all people manually upgrading their machines I strongly recommend to use one of the newest available Clients. However we do have some major contributors that still stick to a 5.8 client, and apparently upgrading is not easy for them.

BM
Profile Gundolf Jahn
Send message
Joined: 1 Mar 05
Posts: 1079
Credit: 341,280
RAC: 0
Message 107886 - Posted: 17 Nov 2010, 11:47:48 UTC - in response to Message 107883.

But why stop at version 5.8?

BOINC 5.8.16 is the last version compatible with windows NT4!

Gruß,
Gundolf
Odysseus
Avatar
Send message
Joined: 17 Dec 05
Posts: 372
Credit: 5,859,232
RAC: 857
Message 107974 - Posted: 22 Nov 2010, 1:37:44 UTC - in response to Message 107886.

I have been unsuccessful getting BOINC v6.x to run on my older Macs as well. Not all of them are yet at v5.8 or higher, but I expect upgrades to that level will be feasible.

____________

Richard Haselgrove
Send message
Joined: 10 Dec 05
Posts: 1721
Credit: 64,780,904
RAC: 56,028
Message 107979 - Posted: 22 Nov 2010, 10:42:51 UTC
Last modified: 22 Nov 2010, 10:43:17 UTC

I've noticed a small problem with this when setting up host 3674581 for a burn-in test.

It's modern server hardware, so should perform well on Einstein tasks. But it's running Windows Small Business Server 2008 - which means it is running (has to run) as a Domain Controller. That in turn means that I'm restricted to BOINC v5.10.xx maximum. And that in turn means that all that lovely hardware is running base-level stock code, with no SIMD optimisations. Seems a shame, really.

mikey
Avatar
Send message
Joined: 22 Jan 05
Posts: 2573
Credit: 160,453,930
RAC: 127,246
Message 107981 - Posted: 22 Nov 2010, 12:14:29 UTC - in response to Message 107886.

But why stop at version 5.8?

BOINC 5.8.16 is the last version compatible with windows NT4!

Gruß,
Gundolf


WOW sooo many things to consider that I never had! Thanks for the info guys!
Brian Priebe
Send message
Joined: 28 Nov 09
Posts: 2
Credit: 25,541,276
RAC: 8
Message 108062 - Posted: 26 Nov 2010, 4:11:10 UTC - in response to Message 107979.
Last modified: 26 Nov 2010, 4:12:29 UTC

But it's running Windows Small Business Server 2008 - which means it is running (has to run) as a Domain Controller. That in turn means that I'm restricted to BOINC v5.10.xx maximum.

Check out this thread over at Collatz: http://boinc.thesonntags.com/collatz/forum_thread.php?id=455&nowrap=true#7137. Thanks to Crunch3r's method, I'm running 6.10.56 on a Windows 2003 Server domain controller under the Administrator account.
Olaf
Send message
Joined: 16 Sep 06
Posts: 25
Credit: 178,078,586
RAC: 58,246
Message 108460 - Posted: 9 Dec 2010, 9:55:50 UTC - in response to Message 107883.


But why stop at version 5.8? Why not make people go all the way to 6.x instead?


My observation with 6.x is, that it fails to work in a more complex environment
than a single computer.
You need an own home directory on any computer to run it on more than one
computer of a linux-cluster, that shares accounts and home directories.
In such a construction you can run one computer with 6.x, the other with 5.x
or stop BOINC on them - at least I found no other simple solution without
generating superfluous additional accounts. If someone knows, how to optimise
6.x to use only information in an arbitrary subdirectory, this would be an
option for me to update BOINC from 5.10 to 6.x. Luckily there seems to be no
urgent problem with 5.10 ;o)
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3611
Credit: 128,377,897
RAC: 55,563
Message 108480 - Posted: 10 Dec 2010, 11:35:29 UTC - in response to Message 108460.
Last modified: 10 Dec 2010, 11:35:52 UTC

Doesn't running the Client without sandboxing (i.e. with command-line option --insecure) help?

BM

Richard Haselgrove
Send message
Joined: 10 Dec 05
Posts: 1721
Credit: 64,780,904
RAC: 56,028
Message 108505 - Posted: 12 Dec 2010, 0:39:19 UTC

I'm pleased to note that Bernd has implemented some improvements in the Einstein server scheduler.

With my BOINC v5.10 installation (highest permitted on a Windows Domain Controller), the new scheduler has issued the SSE2 application, instead of the plain-vanilla app I was running before. This has cut down the running time from ~8 hours with the plain vanilla Global Correlations S5 HF search application, to ~5.5 hours with the SSE2 optimised version.

If you are already running S5GC1HF under BOINC v5.10 or earlier, you won't get the optimised version automatically until the next version upgrade on the server. But if you perform a project reset, you should get it straight away. Note that resetting the project discards all work in progress or ready to report: only do it at a time when you have minimal work on hand, or after you have drained your cache by setting 'No New Tasks'.

Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3611
Credit: 128,377,897
RAC: 55,563
Message 108513 - Posted: 12 Dec 2010, 6:28:04 UTC - in response to Message 108460.

If someone knows, how to optimise
6.x to use only information in an arbitrary subdirectory, this would be an
option for me to update BOINC from 5.10 to 6.x


There are two command-line options of the client that might help:

* "--insecure" should switch off the "sandboxing" (requiring a "boinc" user account)

* "--dir /var/tmp/BOINC" allows to put the BOINC data directory anywhere you like, in the example to /var/tmp/BOINC, which should be local on all machines on a cluster (as an Admin I'd prefer /var/lib/boinc)

BM
Profile Gundolf Jahn
Send message
Joined: 1 Mar 05
Posts: 1079
Credit: 341,280
RAC: 0
Message 108516 - Posted: 12 Dec 2010, 8:38:50 UTC - in response to Message 108505.

If you are already running S5GC1HF under BOINC v5.10 or earlier, you won't get the optimised version automatically until the next version upgrade on the server. But if you perform a project reset, you should get it straight away...

I did get the SSE2 version without resetting the project. :-D
I didn't even notice the change until I'd read your post. :-)

However, my Einstein cache had been empty before the download of einstein_S5GC1HF_3.06_windows_intelx86__S5GCESSE2.exe took place (due to debt issues with SETI I reckon).

Gruß,
Gundolf
Richard Haselgrove
Send message
Joined: 10 Dec 05
Posts: 1721
Credit: 64,780,904
RAC: 56,028
Message 108521 - Posted: 12 Dec 2010, 9:53:49 UTC - in response to Message 108516.

If you are already running S5GC1HF under BOINC v5.10 or earlier, you won't get the optimised version automatically until the next version upgrade on the server. But if you perform a project reset, you should get it straight away...

I did get the SSE2 version without resetting the project. :-D
I didn't even notice the change until I'd read your post. :-)

However, my Einstein cache had been empty before the download of einstein_S5GC1HF_3.06_windows_intelx86__S5GCESSE2.exe took place (due to debt issues with SETI I reckon).

Gruß,
Gundolf

If you haven't downloaded any new work since the x.06 applications were installed on 8 December, yes, you'll get the SSE2 (or whichever version is appropriate) automatically with your next download. But if you have downloaded work since that date, you'll already have an x.06 app, but you won't get upgraded to SSE/SSE2 without a reset or until a new version is installed on the server.
Profile Gundolf Jahn
Send message
Joined: 1 Mar 05
Posts: 1079
Credit: 341,280
RAC: 0
Message 108522 - Posted: 12 Dec 2010, 11:40:03 UTC - in response to Message 108521.
Last modified: 12 Dec 2010, 11:40:28 UTC

Oops, I stand corrected.

I did check the release date of the application but only assumed I would have downloaded some work between then (12/8) and the download of the SSE2 application (12/11). However, my host seems to be slower than I've thought: the latest previous download had occured on 12/5.

Gruß,
Gundolf

juergen.mell
Send message
Joined: 9 Feb 05
Posts: 9
Credit: 1,809,928
RAC: 1,245
Message 108586 - Posted: 15 Dec 2010, 20:08:10 UTC - in response to Message 107883.

AFAIK, version 6 still does not run on a Windows domain controller. Until this is fixed, version 5 should still be supported!
JM
____________

Richard Haselgrove
Send message
Joined: 10 Dec 05
Posts: 1721
Credit: 64,780,904
RAC: 56,028
Message 108587 - Posted: 15 Dec 2010, 20:25:45 UTC - in response to Message 108586.

AFAIK, version 6 still does not run on a Windows domain controller. Until this is fixed, version 5 should still be supported!

Actually, BOINC v6 cannot be installed on a domain controller - it's the installation process with the normal Berkeley installer which fails: there are reports that it runs OK if you install it manually.

But quite a bit of this and the related threads have been about that problem: and Bernd has stated that he has no intention of raising the minimum BOINC version beyond v5.8, and has put considerable effort to ensure that BOINC v5 clients run as efficiently as possible.
1 · 2 · Next

Message boards : Technical News : Raise of minimum required BOINC Core Client version


Home · Your account · Message boards

This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2016 Bruce Allen