i had discusses their benchmark results already with a embarcadero
employee in a german forum
and they insisted that tpc was bad with firebird and good with interbase.
When i reproduced their description of the test on a known powerful
firebird machine
and exactly the same hardware and same OS with Interbase, the results
were not
really different.
All setup changes they tried to make a different result better for
interbase were simply
done in the focus to have options used not available in Firebird or
simply in other ways
implemented in Firebird.
But a real world simple Benchmark like the one built in IBExpert full
version still runs
on Interbase and firebird and this simply creates a enw database from
scratch based
on a script, then creates single threaded test data, do some stupid
things for testing
tmp speed on disk, then runs 10 parallel threads to create new test data
on top of
existing data. all in all it does about 1 million
insert/update/delete/select statements
and runs full job one time on a database with 50 cache buffers (to have
aprox
500000 page read/writes to see hardware hdd/ssd iops performance) and again
with 5000 cache buffers (where result only needs about 3000 page read/writes
to see mainboard/ram/cpu max speed). Total benchmark I/O does 13 GB or 43 GB
read/write on disk depending on architecture (ss or sc/cs)
When using specific params in the benchmark, it will do the same job
with 50 threads
and a 5 time bigger db and an internal version also allows 250 threads
with 25 times
bigger db.
This simple test gives anyone who is using Firebird or Interbase a much
better real world
comparison between different hardware platforms with the same fb or ib
version, but
it gives also a very good and reliable result for real world operations
(when our benchmark
is 5 times faster, long sql statements often also run at least 5 times
faster and daily
jobs like backup and especially restore of run even faster).
But this can also be used to compare interbase and firebird on the sam
ehardware, which
i did several times before and after the above mentioned discussion. And
results were
really easy: Interbase used for the benchmark on identical systems in
default configuration
around twice as long as Firebird needed.
So whatever they are trying to tell the people, they simply bring
misinformations and forget
what for example firebird can do and they are not even able to do it
slowly.
My favourite example: try to copy data from one database to another
incl. blobs.
There is no built in method in Interbase that allows this and any client
based external way
to do so is horrible slow. Try the same with a simple execute block with
execute statement on
external and params, you will be very impressed how fast this is and
compared to external
clients, my experience that it is between 5 and 20 times faster and
still has full commit/rollback
functionality.
We did several jobs for customer to convert interbase databases to
firebird and anytime
i have to work again in Interbase, i miss so many day by day used
functionality i have in
Firebird since years, so whatever they try to tell, their time is
already over and similar as in the
delphi world, they try to squeeze all money out of their existing
customers for some
buzzwords in new version (and i do not report about severe crashes of
interbase databases
which were for most (now satisfied firebird) customers the reason to
switch from
interbase to firebird.
Mit freundlichen Grüßen / Kind regards
Holger Klemt
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel
