Distributed Database Issues

Hi all,

I'm sitting here looking at the white-paper on Distributed Database Technology and Symmetric Replication from the Oracle WWW-site (http://www.oracle.com/info/products/entserver72/ddtsr.html).

I'm wondering about some of the statements Oracle makes in this paper and would be interested in your comments on them.

In the section on "Synchronous Distributed Technology" they say:

"Synchronous technology ensures application integrity and minimizes
complexity, but can have poorer response time performance and less availability if the systems and networks involved are unreliable and slow."

and in the section on "Asynchronous Distributed Technology" they say:

"Asynchronous technology maximizes availability and response time
performance, but can be more complex and requires careful planning and design to enure application integrity."

I'm reading (or interpreting) this (and the rest of the document) to mean that Oracle thinks that using PL/SQL, triggers and procedures is easier to set up and manage than snapshots.

I would think that setting up a synchronous distribution environment would be just as complex as setting up an asynchronous one.

More important, I would think that maintenance and expansion of such an environment would be easier using asynchronous replication (snapshots) than using synchronous distribution (i.e PL/SQL, triggers and procedures). One should think that adding one more site with synchronous distribution would involve writing or re-writing lots of PL/SQL, while adding another snapshot site would be easier (since its deklarative).

Having little experience with this technology, I might be dead wrong. However, it seems to me that if one is using replication, it would be easier to set up and manage snapshots then writing the replication code yourself. I'm most concerned about management and maintenance of such an environment. Setting it up migth be difficult, but it is an onetime job, whilst one could imagine such a system expanding and beeing in operation for years. Thats when it needs to be managable and maintainable.

What are your thoughts?

I'm also wondering if anyone can tell me about the underlying technology for the "Multiple Master" replication they are talking about. They are saying that it uses "Deferred RPC" as the undelying technology, but does that mean that you'll have to write these RPC's yourself? Or is there a layer "over" this that you use (as with snapshots)?

Thanks,
Haakon

-- 
hts_at_sasdata.no | haakon.soenderland_at_thcave.bbs.no | Haakon T. Soenderland
Scandinavian Airlines Data Norway A/S  
---


"40 skiver og et herpa anlegg, en leilighet som trenger aa spyles. Ingen 


 venner og ingen penger, alt jeg har er mine klamme hender.." Jokke '94
Windows 95: A another first from Microsoft!
The first OS to be obsolete *before* it was released.