Thanks for the detailed reply. Will try it out.

On Sun, Feb 13, 2011 at 5:08 AM, Hugh Williams <[email protected]>wrote:

> Hi Abhi,
>
> Although O_DIRECT can be enabled for those OS'es that support it, testing
> performed by development has shown no evidence of performance improvements
> hosting large databases on any of the OS;es  tested on.
>
> If testing on Linux their is a kernel configuration called "swappiness"
> that controls how much the kernel favors swap over RAM:
>
>
> http://www.linuxvox.com/2009/10/what-is-the-linux-kernel-parameter-vm-swappiness/
>
> When hosting large datasets, as you are, it is recommended this parameter
> be changed from is default value of 60 to something like 10, which is what
> we use on the dbpedia and lod etc services we host, to reduce that amount of
> swapping that takes please on the server.
>
> Note:
>
> 1. The swappiness setting resides in the file "/proc/sys/vm/swappiness"
> 2. The command "/sbin/sysctl vm.swappiness" can be used to view its setting
>
> 3. The command "/sbin/sysctl -w vm.swappiness=10" can be used to change its
> value
> 4. Adding the "vm.swappiness = 10" to the file "/etc/sysctl.conf" will
> force the value to be set at machine boot time.
>
> Let me know if his helps.
>
> Best Regards
> Hugh Williams
> Professional Services
> OpenLink Software
> Web: http://www.openlinksw.com
> Support: http://support.openlinksw.com
> Forums: http://boards.openlinksw.com/support
> Twitter: http://twitter.com/OpenLink
>
> On 9 Feb 2011, at 06:18, Abhi wrote:
>
> Sorry for the delayed response, I am not very clear as to how o_direct
> works. If I am not mistaken, I think it has to do something with OS page
> caching, where virtuoso also caches data and OS also caches data in the form
> of page files and this double caching is avoided with the o_direct
> parameter. I maybe completely wrong here. If I am please correct me.
>
> I want to test the performance benefits of using this parameter as we are
> expecting load sizes of billions of triples in our store. All I know is that
> o_direct param in the ini file enables this and this is OS dependent. But,
> is there any document out there which says which OSes support this parameter
> and are there any changes that needs to be done in Linux kernal(while
> compiling the Kernal or some config file) to enable this? Any document on
> this(Unix related or Virtuosos related) will help me in better understanding
> what is going on behind the scene.
>
> On Sun, Jan 23, 2011 at 7:41 AM, Hugh Williams 
> <[email protected]>wrote:
>
>> Hi Abhi,
>>
>> Is this a Unix or Virtuoso question ie do you want to know what OS'es
>> support  it or on what OS'es Virtuoso supports it ?
>>
>> Best Regards
>> Hugh Williams
>> Professional Services
>> OpenLink Software
>> Web: http://www.openlinksw.com
>> Support: http://support.openlinksw.com
>> Forums: http://boards.openlinksw.com/support
>> Twitter: http://twitter.com/OpenLink
>>
>> On 21 Jan 2011, at 03:11, Abhi wrote:
>>
>> > Is there a list of OSes that support the O_DIRECT param or does anybody
>> know how do i find whether a particular OS supports it?
>> >
>> > --
>> > Cheers,
>> > Abhi
>> >
>> ------------------------------------------------------------------------------
>> > Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)!
>> > Finally, a world-class log management solution at an even better
>> price-free!
>> > Download using promo code Free_Logger_4_Dev2Dev. Offer expires
>> > February 28th, so secure your free ArcSight Logger TODAY!
>> >
>> http://p.sf.net/sfu/arcsight-sfd2d_______________________________________________
>> > Virtuoso-users mailing list
>> > [email protected]
>> > https://lists.sourceforge.net/lists/listinfo/virtuoso-users
>>
>>
>
>
> --
> Cheers,
> Abhi
> ------------------------------------------------------------------------------
> The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE:
> Pinpoint memory and threading errors before they happen.
> Find and fix more than 250 security defects in the development cycle.
> Locate bottlenecks in serial and parallel code that limit performance.
>
> http://p.sf.net/sfu/intel-dev2devfeb_______________________________________________
>
> Virtuoso-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/virtuoso-users
>
>
>


-- 
Cheers,
Abhi

Reply via email to