Leon Mergen wrote:
> I recently read an article at Zend.com, showing that Postgres had
better
> performance than MySQL... Off course, this is a very bad thing to say
in the
> MySQL mailing list, and I'm not really sure about it either, since so
many
> praise MySQL...
>
> http://zend.com/zend/art/databases.php
>
First and most important point!
If you follow the link to the news release you will find that the
PostgreSQL people
paid for this
test themselves....hmm remind you of another famous benchmark....
Benchmarks should always be taken with a grain of salt especially when
paid for by one
of the
involved parties. Also, benchmarks are often VERY limited in what they
test. Any
information
gained must be carefully gleaned and the test parameters scrutinized for
unintended
effects.
Database testing especially should have knowledgeable people for
each contender.
Zend admits
that it only uses PostgreSQL, has little MySQL experience and NO
Interbase familiarity.
If you read the comments of the engineer at Xperts Inc., you will see
that his
experience is lacking
with these others as well.
I noticed some significant items in the description of the benchmark
testing.
1) Interbase could not be tested in the TPC-C test because they couldn't
find an ODBC
driver
for it. They are using ODBC connections !@$!?? Sorry, but it is
difficult to find a
slower way
to connect to databases. Also there are ODBC drivers AND there are
ODBC drivers.
They
all connect but performance varies widely between the various
offerings. This alone
pretty
much invalidates this benchmark until it is determined which drivers
were used and
how.
Why not try native connections?
2)TPC-C stands for TRANSACTION Processing Performance Council. Why don't
I choose to
benchmark something that PostgreSQL can't do and castigate it, like
for example,
its poor
BLOB handling. What part of the ANSI spec in MySQL is missing to
remove it from
this test
..transaction processing?
It would be really interested to see how well MySQL and Interbase
would have fared
if they
hadn't been removed from contention before they could beat up on
PostgreSQL, or
would
have they? I guess we will never know.
I think that the line "not adequately compliant with minimal ANSI
SQL Standards"
betrays
something of a bias. This is readily apparent when they discuss the
possibility of
tweaking the
test code for Interbase to get it to run, but not for MySQL.
If you read a comparison of the two, PostgreSQL is missing some
transaction
process
and function capabilities itself. It is not 'Mostly' compliant as
the Zend page
would have you believe.
Check a neutral page like the 'crashme' comparison on www.mysql.com
<g>.
Seriously, which ANSI spec is not met by MySQL that disallows it? I
agree that the
test would
not be fair because PostgreSQL has all that Transaction overhead and
would be
clobbered.
MySQL apps can be transaction safe, if written carefully, and
remember Transactions
have their
faults too. They are not a silver bullet, only a tool with
trade-offs.
More importantly, MySQL does support transactions and row level
locking, etc.,
if you use the
other table types ( Berkeley, Gemini, Inno, etal ).
3)Tuning.
They tuned PostgreSQL because it couldn't do the TPC-C and AS3AP
tests 'out of the
box'
but then they INCREASED the size of the cache too, because it's
resource
conservative! They
also did a 'vacuum analyze' of tables and indexes.
They admit that MySQL and Interbase were used as is.
What good is a benchmark comparison that will not even run on two of the
three
contenders on
50% of the testing? It is simply garbage data and a smokescreen worthy
of M$.
WHY..
did they not chose something that worked on all five contenders?
did they tweak any of the DB's rather than only some? Without
tweaking NONE of the
contenders would have run on ANY of the tests...
ODBC only, rather than native as well?
I am not a MySQL zealot. It has its place, its faults and its good
points. Most major
DB's have a
place in the sun where they rule. It is a matter of design and
evaluation.
I just dislike to see self interested commercialism masquerade as
science. I still
would like to see
a good comparison between these DBs, because, to date I have still not
seen a good one.
danh
---------------------------------------------------------------------
Before posting, please check:
http://www.mysql.com/manual.php (the manual)
http://lists.mysql.com/ (the list archive)
To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
