[BlueBream] how powerfull is ZODB?!

[Justin Ryan]

First..

Noticed this was a private response.. Copying the list ..

Also, as for ZODB and Mounts, splitting storage, etc..

Again, this is something that Z3/ZTK/BlueBream has always lacked afaik.

There is a Z2 product, maybe this would be a great thing for a
fundable or kickstarter project, to award some PayPal rewards to
successful implementers after an open in-person / IRC sprint.

It shouldn't be that much trouble to implement a ZODBMountPoint type
in pure ZTK, after reseaching the Z2 product.

This is on my list, but I'd love to just organize a sprint on it.  I
keep telling customers it's possible, but needs to be funded.  I'd
love to just say, hey, we're doing it.

In Plone, it's common to use mount points for the catalog, as this is
a very basic database strategy.  In PostgreSQL, Oracle, MS-SQL, IBM
DB2, MaxDB, etc.. it's a very basic concept to take a separate
physical device, RAID, whatever, and put the indexes on it.

Same goes for any type of data segmentation you can do.

ZODB Mount points are great for that and it's just sad we don't
currently have a publicly available working solution for them, afaik,
in ZTK/BB.

On Sun, Jul 4, 2010 at 4:10 PM, Justin Ryan
<justin.ryan at reliefgarden.org> wrote:
>> This is the question, in how far I should splitt it up. Perhaps I have
>> to search for something somewhere.... (Propper books). Because if I
>> really would use ZODB, the last thing to use everything in a single file.
>> Perhaps 200 MB for each file (that would mean 1000 files ?!)
>
> I've dealt with many-GB ZODBs with thousands of objects.
>
> With careful expertise, I've heard of ZODBs with millions of objects,
> in Plone, which is at least order of magnitude less efficient than
> basic BB content.
>
> I recently ran a benchmark of an application with something like 1000
> 1kb comments using nested ZTK container which is the worst case, for
> 10x the number of apps in Apple's App Store, so aroudn 2.5M.
>
> I got to about 1M objects before I got bored, but it was only as big
> of a ZODB as a Plone site full of photos and/or PDFs, and content
> addressed by URL (as found via RSS, for instance) loaded very quickly.
>
>>>
>> Yes, you are entirely right. No need to design tables with relations,
>> and later "the ahhhhhhh ughhhh, if i had done that and this ....."
>
> Well, I just say keep it simple.
>
> You can also proxy IContainer implementations to external storage like
> Cassandra or Tokyo Tyrant.
>
> There are a million ways to do things.  ZODB can be helpful with all
> of them I've seen.  Even as just a very small <1MB db for application
> configuration with an RDBMS app, it's pretty handy.
>
>> SAP MaxDB was my 1st choice (also free of charge)
>
> MaxDB has its' ups and downs, it's a fine RDBMS as RDBMS go.  IBM DB2
> I've seen outperform in some corner cases, but both beat Oracle in a
> 10-vendor shootout circa 2004.
>
> ZODB can do a lot of things, but you need to want to learn it.  If you
> want to use an RDBMS, you can still take care of a ton of BlueBream
> advantages.
>
> I only advocate as a successful user, trying not to be a zealot.
>