Object-Persistence in database -- kind of data

[Ruslan Zasukhin]

On 12/15/05 9:25 PM, "Philip Mötteli" <philip.moetteli at econophone.ch> wrote:

>> I still believe that such layer
> 
> You mean object-relational persistency mapper?
> 
>> can be used only for something simple. But I do not think it is good idea
>> choose it for db with 50 tables, 100  relations millions records..
> 
> EOF was used by companies with HUGE, HUGE number of data sets.
> Several terabytes, I know of some astronomical researches. Big
> companies, like UBS have used it for with terabytes. And apart from
> that, the Apple Store also uses EOF. And there are and were many,
> many more.

I knew you will say this :-)

When somebody talk about terabytes, I always want ask:
    what hardware? My DUAL G5 has 80Gb HDD.
    so you use some fast RAID probably?
    how much faster? 10 times?
    Aha, then your terra-byte is equivalent to 100Gb already..
    and so on...

Problem is that today on fast computers, really even power dbs can satisfy.
But I always remember about such tasks as AI:
    - you have computer which can think?
    - talk?
    - understand sense ?
    - ... 

No? Why? Slow computers? Millions times slower than needed? Ops... :-)


>> Okay, it may be will work, but normal solution on relational or OR  model
>> will work 10-20-how many times faster?
 
> That depends on what kind of data you have:

> Data, that is organized  in a complex structure, of a very interconnected
> graph, is better off  with a OODBMS.
> 
> The OODBMS only follows the links. There's nothing  more direct and easy. An
> RDBMS wouldn't be slower here.
> 
> Data, that has to be searched through its contents, is better off in  a pure
> RDBMS.

Ah-ha, so you perfectly understand this. Good.

Yes, if task is ONLY
    go from one object to another, load it as whole
    then go to next object and load it as whole ...

Then right, storing of object as atom is best.

But where exists such tasks ???

We always need search:
    find car fith such number
    and such color and owner is man.

Storing of objects as WHOLE, not-splinted into tables cause a lots of
problems here. I talk about speed. It is even hard to index such mess.

Let me tell you small secret of Valentina (it is documented actually :).
Valentina not just keep data normalized as do RDBMS, Valentina even keep
each column as separate logical file. In many tasks this give great speedup.
And this is very perfect for inheritance of tables, which I hope we will get
sometimes. Because look, data of T1 and T2 do not differ practically in
contrast to solid tables as have MS, Oracle, mySQL, Postgre, ...
 
>> In other words, his db is CLOSED for use in other language, PHP for
>> example?
> 
> No, the data is accessible from every language, which has an
> interface written for the OODBMS this specific DB is living on. It is
> perfectly possible, that a Java, an ObjC, a Perl and a PHP program
> simultaneously access the same data on the same DB.

Yes, agree. I have also think about this after send letter.

Such examples exists? How Perl, PHP can simulate access to OO layer?
In some ugly way I guess ?


-- 
Best regards,

Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc

Valentina - Joining Worlds of Information
http://www.paradigmasoft.com

[I feel the need: the need for speed]