increasing order of oids?

[Jürgen Herrmann]

hi there!

is there a strict requirement for oids to be increasing? having read
a ton of code my head is not so sure anymore, though i can't remember
seeing a requirement for such anywhere in the code.

if RadosStorage could pre-allocate oids in batches of thousands, that
would probably give me a big performance win, or to put it the other
way round: rados is not designed to be used for such time-critical and
at the same time completely synced operations. and getting new oids
will happen often...

best regards,
jürgen
--
>> XLhost.de - eXperts in Linux hosting ® <<

XLhost.de GmbH
Jürgen Herrmann, Geschäftsführer
Boelckestrasse 21, 93051 Regensburg, Germany

Geschäftsführer: Volker Geith, Jürgen Herrmann
Registriert unter: HRB9918
Umsatzsteuer-Identifikationsnummer: DE245931218

Fon:  +49 (0)800 XLHOSTDE [0800 95467833]
Fax:  [phone]

WEB:  http://www.XLhost.de
IRC:  #XLhost@...

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

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

[Shane Hathaway]

Jürgen Herrmann wrote:
> is there a strict requirement for oids to be increasing?

No, but obviously there is a strict requirement that new_oid() return
unique OIDs.

> if RadosStorage could pre-allocate oids in batches of thousands, that
> would probably give me a big performance win, or to put it the other
> way round: rados is not designed to be used for such time-critical and
> at the same time completely synced operations. and getting new oids
> will happen often...

ZEO also allocates OIDs in large batches, I believe.

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

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

[Jim Fulton]

On Thu, Aug 13, 2009 at 2:01 PM, Shane Hathaway<shane@...> wrote:

> Jürgen Herrmann wrote:
>> is there a strict requirement for oids to be increasing?
>
> No, but obviously there is a strict requirement that new_oid() return
> unique OIDs.
>
>> if RadosStorage could pre-allocate oids in batches of thousands, that
>> would probably give me a big performance win, or to put it the other
>> way round: rados is not designed to be used for such time-critical and
>> at the same time completely synced operations. and getting new oids
>> will happen often...
>
> ZEO also allocates OIDs in large batches, I believe.

Yup.  Note that technically, IODs can be any string, however lots of
tools assume they are 8-byte strings, which gives plenty of room for
wastage when allocating them.

Jim

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

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