Re: [Pbxt-discuss] Filesystem-cache and PBXT

[Paul McCullagh]

Hi Mark, Erkan

I don't know if I can answer all questions myself.

But the latest version of 1.0 and 1.1 do not use memory mapping by  
default. So I/O is conventional.


On Mar 19, 2010, at 6:47 PM, MARK CALLAGHAN wrote:


Me too. I need to learn more about the different files used by PBXT.
What access patterns it uses for the (random, sequential), whether
they use mmap, when are they forced to disk, ...

I have lots of things to learn.

On Fri, Mar 19, 2010 at 10:36 AM, erkan yanar erkan.ya...@linsenraum.de 
 wrote:


Moin Mark,

reading your post I actively ignored the tmpfs-thing, sorry.
The general question about the way PBXT uses file-system-cache still
stays (for me).


erkan


btw: It hard to comment if not in facebook:)

On Fri, Mar 19, 2010 at 10:17:00AM -0700, MARK CALLAGHAN wrote:

I don't think so. I think all engines were dependent to a similar
degree on tmpfs and nothing go to use the OS buffer cache.

AFAIK, tmpfs files are not doubled-buffered in the OS buffer  
cache. I

think I learned this from reading lkml, but I don't mind citations
that state otherwise.

So, each fetch of a MyISAM row must get all columns and read from  
the
file after doing the primary key index lookup that is likely to  
hit in

the key cache.

InnoDB is should do reads from tmpfs for most queries as the table
data doesn't fit in the InnoDB buffer cache.

I only set two of the my.cnf settings for PBXT and don't know if  
that

was sufficient to keep it from caching too much.

On Fri, Mar 19, 2010 at 9:13 AM, erkan yanar erkan.ya...@linsenraum.de 
 wrote:


Moin,

I enjoy it seeing PBXT gaining more publicity \o/
(http://www.facebook.com/note.php?note_id=378043115932)

I wonder if PBXT benefits as much as MyISAM from filesystemcache?
Reading the post I think MyISAM is better configured on that  
machine
than InnoDB, because MyISAM benefits a lot from the  
filesystemcache.

And now I wonder about best practice for PBXT.


regards
erkan




--
über den grenzen muß die freiheit wohl wolkenlos sein

___
Mailing list: https://launchpad.net/~pbxt-discuss
Post to : pbxt-discuss@lists.launchpad.net
Unsubscribe : https://launchpad.net/~pbxt-discuss
More help   : https://help.launchpad.net/ListHelp





--
Mark Callaghan
mdcal...@gmail.com


--
über den grenzen muß die freiheit wohl wolkenlos sein





--
Mark Callaghan
mdcal...@gmail.com

___
Mailing list: https://launchpad.net/~pbxt-discuss
Post to : pbxt-discuss@lists.launchpad.net
Unsubscribe : https://launchpad.net/~pbxt-discuss
More help   : https://help.launchpad.net/ListHelp




--
Paul McCullagh
PrimeBase Technologies
www.primebase.org
www.blobstreaming.org
pbxt.blogspot.com




___
Mailing list: https://launchpad.net/~pbxt-discuss
Post to : pbxt-discuss@lists.launchpad.net
Unsubscribe : https://launchpad.net/~pbxt-discuss
More help   : https://help.launchpad.net/ListHelp

Re: [Pbxt-discuss] Filesystem-cache and PBXT

[Vladimir Kolesnikov]

Hi,

afaik, there's no specific technique used to utilize the fact that a 
record might be in FS cache like cheap random access or something... 
Another thing - does PBXT depend on cheap random access at all. I used 
blkstat on DBT2 and it appears that there's still some room for 
improvement (e.g. for .xtd file write pattern) but in general most of 
I/O is sequential. Also this will depend on whether you have fixed 
format rows.


BR,
Vlad

03.2010 18:13, erkan yanar пишет:

Moin,

I enjoy it seeing PBXT gaining more publicity \o/
(http://www.facebook.com/note.php?note_id=378043115932)

I wonder if PBXT benefits as much as MyISAM from filesystemcache?
Reading the post I think MyISAM is better configured on that machine
than InnoDB, because MyISAM benefits a lot from the filesystemcache.
And now I wonder about best practice for PBXT.


regards
erkan




   



___
Mailing list: https://launchpad.net/~pbxt-discuss
Post to : pbxt-discuss@lists.launchpad.net
Unsubscribe : https://launchpad.net/~pbxt-discuss
More help   : https://help.launchpad.net/ListHelp