[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: PDOM implementation



1. Transactions

IPSI PDOM have only write transaction ...

it modifies pages of nodes in memory,
and commit method - flushes only modified pages back to disk.

They don't have read-only transactions,
so when one thread modifies in memory pages,
other thread can read it before commit ...

2. Paging / Clustering

Falko, is this possible to store Pages of Nodes instead of single Nodes ?
Something similar to BLOB's implementation in Ozone ...

Falko Braeutigam wrote:

> On Wed, 27 Oct 1999, Ann Tecklenburg wrote:
> > http://xml.darmstadt.gmd.de/xql/apidoc/index.html
> >
> > http://xml.darmstadt.gmd.de/xql/xql-examples.html#PDOM
> >
> > There is really alot of detail about how they implemented their
> > PDOM in the last reference although no source code.
> >
> > I think  their implementation represents an educated and
> > fairly efficient approach.  We will have to do much the same
> > in Ozone to ever get decent performance.
> IPSIs PDOM is lightning fast compared against what we we can do today. But the
> goals and strategies behind seem to be very different from our MDOM so I don't
> know, if we can re-use that much of their "knowledge".
>
> I see two ways to implement a persistent DOM:
> 1) the DOM classes know how to deal with persistence (IPSI PDOM)
> 2) the DOM nodes are stored in a database (ozoneXML MDOM)
>
> The PDOM solution is fast because all the DOM-specific information can be used
> to optimize the storage (the DOM code itself includes the storage code) and it
> does not need any server. But this is also the problem, as far as I see, it
> is impossible to build a transaction save system with PDOM. (a feature that
> is a must for a document server for example) And if they start to use a database
> as storage backend to come around this problem, they will lose the performance
> advantages.
>
> In the future we will try to re-work especially the cluster management of ozone
> to better meet the needs of DOM. The performance of ozoneXML must be
> comparable to PDOM. However, it will never be as efficient as such an
> specialized system.
>
> Falko
> --
> ______________________________________________________________________
> Falko Braeutigam                         mailto:falko@softwarebuero.de
> softwarebuero m&b (SMB)                    http://www.softwarebuero.de

--
-------------------------------------------------
Zvi Avraham, Senior Software Engineer
NetManage Inc., Visual Connectivity Division
http://www.netmanage.com/products/visual_conn.asp