RE: [Zope] DBTab product questions (more...)
Performance-wise, wouldn't partitioning your storages have the _potential_ to be a win if you were able to distribute transactions onto different ClientStorages on different dedicated ZSS nodes, assuming heavy usage? At what point does one want to worry about the ZSS as a performance bottleneck (or should one worry at all)? I've always thought partitioning via multiple, mounted storages would be the answer, but I've never known enough to confirm or deny this? Sean -----Original Message----- From: Shane Hathaway [mailto:shane@zope.com] Sent: Wednesday, March 19, 2003 9:59 AM To: Thierry FLORAC Cc: zope@zope.org Subject: Re: [Zope] DBTab product questions (more...) On Wed, 19 Mar 2003, Thierry FLORAC wrote:
I just have another little question : my old development database had a size of 2.8 MB ; I've "extracted" my ZCatalog from it to a new ZODB, and tried to pack the two databases : now, each of them has a size of nearly 2.1 or 2.5
MB, which if "far" more than the original database, so my question : why ??
The Control_Panel stores about 2-3 MB of data, and each database has its own copy of the Control_Panel. I wouldn't worry about it though.
And another one : does the fact of splitting a Zope site between several ZODBs can have any impact (improvement or not) on the whole performances ??
There is no measurable difference in performance unless you mount a large number of databases. The MountPoint._setDB() and MountPoint.close() methods are the issue here: they have to open/close all mounted connections, even if they won't be used in the course of a transaction. There are surely ways to improve this. Shane _______________________________________________ Zope maillist - Zope@zope.org http://mail.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://mail.zope.org/mailman/listinfo/zope-announce http://mail.zope.org/mailman/listinfo/zope-dev )
sean.upton@uniontrib.com wrote:
Performance-wise, wouldn't partitioning your storages have the _potential_ to be a win if you were able to distribute transactions onto different ClientStorages on different dedicated ZSS nodes, assuming heavy usage? At what point does one want to worry about the ZSS as a performance bottleneck (or should one worry at all)? I've always thought partitioning via multiple, mounted storages would be the answer, but I've never known enough to confirm or deny this?
Well, FWIW, I think you're correct. ;-) Shane
On Wednesday 19 March 2003 7:03 pm, sean.upton@uniontrib.com wrote:
Performance-wise, wouldn't partitioning your storages have the _potential_ to be a win if you were able to distribute transactions onto different ClientStorages on different dedicated ZSS nodes, assuming heavy usage?
In ZEO1 this limit was actually quite easy to reach with concurrent writes. ZEO2 is better, perhaps 'good enough', but not as good as it could be. I guess any future gains will come from making ZSS multi-threaded, and allow the storage to process concurrent reads. -- Toby Dickenson http://www.geminidataloggers.com/people/tdickenson
participants (3)
-
sean.upton@uniontrib.com -
Shane Hathaway -
Toby Dickenson