[General] Any linux gurus?

[Stephan Henning]

-WD

Ya, I'm going to recommend we look into this and see if other options give
us some improvement


-David

The current benchmark run takes about 10 days to fully solve, which is
about on par with our longest/largest 'production' problem.

The machines are on a UPS, but it is more for power conditioning than
uptime. With three of these machines in the rack running all out, the
largest rackmount units are only good for a few minutes. However, the
program does 'checkpoint' itself at certain intervals, so we will loose
some data with a power loss, but usually no more than a few hours worth on
these really long solves.


On Mon, Dec 16, 2013 at 6:28 PM, David <ainut at knology.net> wrote:

> I had completely forgotten about the overhead of atime and did not know
> the practicality of noatime.  Good catch, Charles.
>
> So I recommend that you use a filesystem on the ssd's that does not use
> journaling, and set the ssd drive directories with (as root) :
>
> format the SSD's with ext2 (fat32 is not recommended for high speed usage
> but may not matter with ssd's and only two files);
>
> then,
>
> [root at machine /root]#*chattr*  -R +A  /dirofSSDtempdata/
> (after reformat, of course.)
>
>
> Caveat: without journaling, ensure you have a good battery backup APS on
> the system.  You probably already have that for something that runs for
> WEEKS!
>
> Also, I hope whoever wrote the program made it re=entrant with little loss
> of time in case the system did crash, for whatever reason.
>
>
> David M.
>
>
>
> Charles Moye wrote:
>
>> and set noatime?
>>
>>
>>
>> On Mon, Dec 16, 2013 at 9:45 AM, David <ainut at knology.net <mailto:
>> ainut at knology.net>> wrote:
>>
>>     Can't remember if ext4 uses journaling but if it does, you'll get
>>     better times if you can turn it off.
>>
>>
>>
>>     Stephan Henning wrote:
>>
>>>     Wow, nice.
>>>
>>>     -WD, I've gone in and checked, and at least on the SSD array,
>>>     it's ext4.
>>>
>>>
>>>     On Mon, Dec 16, 2013 at 7:01 AM, Matt <brimstone at the.narro.ws
>>>     <mailto:brimstone at the.narro.ws>> wrote:
>>>
>>>         On 12/15/2013 10:37 PM, Arthur wrote:
>>>         > What did mailman do?
>>>
>>>         Arthur,
>>>
>>>         Mailman's default setting for digests is to send them out
>>>         every 30KB.
>>>         People don't trim messages like the bad old days; Ethan's
>>>         response email
>>>         after yours is 35KB alone. I upped the setting to 10MB, which
>>>         is Gmail's
>>>         inbound limit if I recall correctly.
>>>
>>>         #matt
>>>
>>>
>>>
>>>         _______________________________________________
>>>         General mailing list
>>>         General at lists.makerslocal.org
>>>         <mailto:General at lists.makerslocal.org>
>>>
>>>         http://lists.makerslocal.org/mailman/listinfo/general
>>>
>>>
>>>
>>>
>>>     _______________________________________________
>>>     General mailing list
>>>     General at lists.makerslocal.org  <mailto:General at lists.makerslocal.org
>>> >
>>>     http://lists.makerslocal.org/mailman/listinfo/general
>>>
>>
>>
>>     _______________________________________________
>>     General mailing list
>>     General at lists.makerslocal.org <mailto:General at lists.makerslocal.org>
>>
>>     http://lists.makerslocal.org/mailman/listinfo/general
>>
>>
>>
>>
>> _______________________________________________
>> General mailing list
>> General at lists.makerslocal.org
>> http://lists.makerslocal.org/mailman/listinfo/general
>>
>
>
> _______________________________________________
> General mailing list
> General at lists.makerslocal.org
> http://lists.makerslocal.org/mailman/listinfo/general
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.makerslocal.org/pipermail/general/attachments/20131216/ee7a92d0/attachment.html>