Greg Stark wrote:
>
> Tom Lane <tgl@sss.pgh.pa.us> writes:
>
> > So? You're going to be paying those costs *anyway*, because most of the
> > process context swaps will be between the application server and the
> > database.
>
> Separating the database and application onto dedicated machines is normally
> the first major optimization busy sites do when they discover that having the
> two on the same machine never scales well.
If there is enough of an "application" to separate it ... :-)
People talk about "database backed websites" when all they need is
thousands of single index lookups. Can someone give me a real world
example of such a website? And if so, what's wrong with using ndbm/gdbm?
All these hypothetical arguments based on "the little test I made" don't
lead to anything. I can create such tests that push the CPU load to 20
or more or get all the drive LED's nervous in no time. They don't tell
anything, that's the problem. They are purely synthetic. They hammer a
few different simple queries in a totally unrealistic, hysteric fashion
against a database and are called benchmarks. There are absolutely no
means of consistency checks built into the tests and if one really runs
checksum tests after 100 concurrent clients hammered this other super
fast superior sql database for 10 minutes people wonder how inconsistent
it can become after 10 minutes ... without a single error message.
Anyone ever thought about a reference implementation of TPC-W?
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #
