Re: Aberdeen Study on OS RDBMS: exceprts and breakdown.

On Wednesday 31 March 2004 21:44, Robert Treat wrote:
> The Aberdeen group recently released a study on open source databases,
> focusing on BerkleyDB, My$QL, and PostgreSQL.

> I would point
> out one item I thought was interesting; they refer to BerkleyDB as "the
> most widely used application-specific data management software in the
> world with over 200 million deployments". I have no reason to believe
> this isn't true, but I do notice that it does rub against my$ql's slogan
> of being "the worlds most popular open source database".

Ah, the difference between "data management software" a database, a DBMS,
RDBMS...

> The following is their general description of PostgreSQL, my comments
> are surrounded by ##
>
> The PostgreSQL Open Source Database
>
> PostgreSQL is a full relational database with SQL support. Many users
> view PostgreSQL as more complex and powerful than MySQL, but much less
> powerful than an enterprise database. Interestingly, some knowledgeable
> MySQL users view MySQL as more scalable than PostgreSQL.
>
> # I think this links back to my$ql's ability to scale up on read only
> type database systems, and having some big names associated with this
> type of use (yahoo for instance). The other half being our legacy of
> being a "slow" database. #

This is what irritates me - looks like information, but in fact is second-hand
gossip. "Many users view X as Y" - well, are they right? Under what
circumstances? When is the situation reversed?

Although it'd be better from an advocacy viewpoint if they said "PG users said
it scales better than MySQL" it wouldn't be any more useful as information.

> PostgreSQL offers the following key features beyond the relational
> “basics”:
>
> * Tablespace backups, i.e., the ability to back up a set of tables
> rather than the entire database — with limits

Um - can we do multiple tables at once?

> * ODBC support
> * Online backup (with limits)

Are they talking about PITR as the limit here?

> PostgreSQL does not offer the following key features typically provided
> in enterprise databases:
>
> * Incremental and parallel backup/restore
> * Encryption (security)
> * Deadlock detection
>
> # but up above you said we did have it... ? I'd grant that it may not
> be as robust as some of the commercial guys, but it's there. #

I'm puzzled as to what they want from encryption - we do SSL on the
connections (don't we?) and you can encrypt the data in specific columns for
passwords etc.
Are they talking about encrypting the actual data blocks as they get written
to disk? If so, I guess you could run on an encrypted filesystem.

> * Bit-mapped indexing (for large data warehouses)
> * A single GUI administrative interface
>
> # again... nothing packaged with the server itself, though we do have
> several options in this area #

We get quite a few requests for a GUI on the lists too. Maybe we need to look
at providing some pointers from the download page.

> * View update/insert/delete
>
> # cough...rules..cough#
>
> PostgreSQL users indicate that improvements in some areas are
> contemplated, but the next version of PostgreSQL may come as late as two
> years from now, and the community that drives PostgreSQL development is
> not committing to any particular improvements.
>
> # uh... two years out... i couldn't come up with an explanation for
> that... well, many of these features are planned for in 7.5, and I
> wonder if it is worth coming up with a "response" to the information
> above #

As Bruce says, the 2 years thing is extremely puzzling. Especially as they
could have just looked at the release notes to see actual dates.

> when asked why the ISV's chose an OS DB "All interviewees immediately
> cited “price” (license cost) as the main reason."

I'm sure they said that, but I think it's actually convenience and
price/performance.

> when asked what they would like to see from their current open source
> vendor... "One interviewee cited full-text indexing, one cited triggers
> and a “geographic information system (GIS),” one cited rollback, and one
> cited a GUI “control center.”"... seems they should all be looking at
> postgresql a little closer :-) Also in response to that question was
> this telling line.. "Two-phase commit is not available in PostgreSQL.
> Every PostgreSQL user requested this feature." ... makes you wish you
> email those folks about the patches that have been posted to -hackers

EVERY PostgreSQL user - all of them? Were they prompted, is there some sort of
conspiracy? Or were they asked to name their "top 100 most desired features".

> finally, of the respondents in the study whom were using proprietary
> databases for the past several years, none expressed interest in trying
> an open source database package.

Not surprising, I suppose. If you asked them whether they'd switch to another
proprietary platform free of charge, they'd probably still say no. Expensive
proposition, switching.

> Hope this was informative for everyone here, and this will hopefully
> help give some inspiration to new ideas on how we can push forward
> postgresql adoption in the future.

Thanks Robert - most interesting.

Should we write an open letter response (need to be carefully pitched though).

--
Richard Huxton
Archonet Ltd