wichert wrote:
> 
> On 2009-9-21 17:38, Martin Aspeli wrote:
>> Maybe so. I've yet to benchmark. If the blob doesn't slow down the
>> "view" use case, then storing the (infrequently-used) raw value in a
>> blob may possibly mean more efficient use of storage and memory. The
>> ZODB cache won't need to hold the full raw value unless it is actually
>> read.
> 
> I am a bit worried that using multiple blobs (one for each text field) 
> for each content item in a large site will result in many many small 
> files, which will cause performance problems at the OS/filesystem level. 
> Not all filesystems can do tail packing which means you will loose a lot 
> of diskspace, and the amount of files can slow down directory operations 
> and fsck.
> 
> Wichert.
> 

I wander if we should write a blob storage or cache using FileStorage. The
whole file or individual blob portions of it could be mmapped to avoid
python allocating memory for long strings of page content. Similar to
Varnish, it would rely on the VFS page cache to make things fast. Multiple
processes could share the same blob cache file, saving memory.

Laurence
-- 
View this message in context: 
http://www.nabble.com/Blob%2C-persistence-and-parentage-questions-tp25529169p25649671.html
Sent from the Zope - ZODB-Dev mailing list archive at Nabble.com.

_______________________________________________
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
https://mail.zope.org/mailman/listinfo/zodb-dev

Reply via email to