Object Databases

[Ng Pheng Siong]

In article <3dk8jnd5un.fsf_-_ at amarok.cnri.reston.va.us>,
  akuchlin at mems-exchange.org (Andrew M. Kuchling) wrote:
> 2) ZODB.  Pro: very natural for Python users; very flexible.
> Downsides: only one database at a time; doubtful scalability once you
> start pushing lots of 5Mb images into it.

Not necessarily one database at a time:

import ZODB
import ZODB.FileStorage

def new_odb(fs):
    conn = ZODB.DB(ZODB.FileStorage.FileStorage(fs)).open()
    get_transaction().commit()
    return conn

def commit():
    get_transaction().setUser('ngps')
    get_transaction().commit()

def put2():
    c1 = new_odb('c1.fs')
    c2 = new_odb('c2.fs')
    r1 = c1.root()
    r2 = c2.root()
    r1['A'] = 'AAA'
    commit()
    r2['a'] = 'aaa'
    commit()
    r1['B'] = 'BBB'
    commit()
    r2['b'] = 'bbb'
    commit()
    c1.close()
    c2.close()

Tranalyzer reports thusly:

/usr/local/home/ngps/zope:$ ./tranalyzer.py c1.fs
TID: 332C81A7229C777 obs 1 len 139 By [Zope]
"initial database creation"
        OID: 0 len 57

TID: 332C81A7234B3AA obs 1 len 140 By / ngps
""
        OID: 0 len 77

TID: 332C81A7234B3AB obs 1 len 151 By / ngps
""
        OID: 0 len 88

****** TRUNCATED TRANSACTION at 315:
TID: 332C81A7234B3AB obs 1 len 151 By / ngps
""
        OID: 0 len 88

I'm too lazy to cook up a multi-threading test program for the
above. Perhaps you may wish to try...?

(I envisage a Zope product that writes to local ZODB, instead of
var/Data.fs. We shall see if that is feasible...)

As noted by Jim Fulton on the Zope list, cross-referencing and
some other issues exist. If your object space partitions nicely,
this might not be that big a problem.

High transaction rates involving 5Mb images will likely cause
difficulties for RDBMS's, too. (The usual tricks involve storing
the raw data out of the DB.) I have insufficient experience
to offer anything beyond that opinion. ;-)

Cheers.

--
Ng Pheng Siong <ngps at post1.com> * http://www.post1.com/home/ngps


Sent via Deja.com http://www.deja.com/
Before you buy.