Fore!

OpenSolaris Build 110 Available - Help Test CD Install Performance

Friday Mar 27, 2009

Late yesterday we released build 110 of the OpenSolaris distro, see [osol-announce] OpenSolaris development package repository build 110, x86/x64/SPARC for the particulars.

Since I was traveling and missed posting a note about build 109, it's probably a good idea to mention one of the changes that appeared there and continues with 110: going to only a single, global CD image.  This had been our intention all along since starting the distribution, but it depended on the LZMA (de)compression support in lofi; we got that done before the 2008.05 release, but its performance was judged to be too slow at the time so we made the expedient decision to ship two different CD's, with the primary CD using the better-performing gzip algorithm.  We figured we'd come back to that issue once we got the LZMA code integrated into the ON gate, which just finally happened a couple of weeks ago.

In the meantime, OpenSolaris community member (and super-contributor!) Juergen Keil took an interest in the performance issue, and found the primary culprit in the lofi code, which was repeated decompression of the same data (see bug 1119 for the discussion if you're curious).  The performance improvement is, we believe, good enough to return to just the single image for the next release.  But we would like data to confirm that, so if you do an install from the build 109 or 110 CD, please send info on the time it took (you can just send the installation log if you like; it's found at /var/sadm/system/logs/install_log after installation) and the basic stats on your hardware (processor speed and number of cores, memory size) to caiman-discuss@opensolaris.org, we'd appreciate it.



[9] Comments
Like this post? del.icio.us | furl | slashdot | technorati | digg
Posted at 10:24AM Mar 27, 2009 by dminer in OpenSolaris_Releases
Tags:
Comments:

Doing it on VirtualBox now from an .iso image just to see. Are you interested in feedback from that as well?

Posted by Michael Ernest on March 27, 2009 at 11:57 AM EDT #

Sure, Virtual Box feedback is welcome; memory configuration of the VM, as well as the CPU and OS info on your host system, would be helpful in that case.

Posted by Dave Miner on March 27, 2009 at 12:12 PM EDT #

Thanks for mentioning Juergen (Keil).

When build 109 first came out, I reported a suspend-to-RAM problem before I went to bed at about 11 pm (Hawaiian time). The next morning when I got to my office at about 7 am (again, Hawaiian time), a workaround was proposed by Juergen, and it fixed the problem.

'nough said!

Posted by W. Wayne Liauh on March 27, 2009 at 01:42 PM EDT #

Is there any chance to have a non-AI build for SPARC? I'd like to give it a try, but honestly, the AI is overkill for me as I have only one spare Sun Blade for testing.

Posted by Petr Vysloužil on March 27, 2009 at 05:21 PM EDT #

Me again. I install os0906_110 in os0906_109 under VBox 2.2-beta2. Everything now seems to be working except the scroll mouse. If anyone can save me the time to dig around for a temporary fix, will appreciate that! :-)

Booting the os0906_110 LiveCD image in VBox took about 3 minutes.

Posted by W. Wayne Liauh on March 27, 2009 at 07:48 PM EDT #

I tested build 110 in Virtualbox on Linux amd64, giving osol 1gb ram 8gb disk, 3d, vt-x, nested paging. The hardware is a 2Ghz core2 duo, with 3gb ram. The install time felt snappier than previous builds. Will try 'bare metal' soon.

One thing though, after install no swap was configured, which is a bit of a problem...

Posted by Andreas Nilsson on March 28, 2009 at 08:13 AM EDT #

Me again. On 'bare metall' the install took ~17 minutes, which I think is a real progress compared to 2008.05. However suspend-to-ram does not want to cooperate, the only thing that happens is that the screen gets locked...

Posted by Andreas Nilsson on March 29, 2009 at 06:17 AM EDT #

Thanks for the results posted.

@Petr: non-AI install options will be developed for future releases, but at this time it's your only choice.

@Andreas: please file bugs at defect.opensolaris.org for the swap and suspend issues you note.

Posted by Dave Miner on March 30, 2009 at 11:43 AM EDT #

@Dave: bugs filed. Hoping for speedy resolving of at least the suspend problem. The virtualbox issue is easily dealt with ( just manualy add a swap device ).

Posted by Andreas Nilsson on April 02, 2009 at 03:05 AM EDT #

Post a Comment:
Comments are closed for this entry.