Re: Is there a meaningful benchmark?

On Fri, Mar 27, 2009 at 2:27 PM, Will Rutherdale (rutherw)
<rutherw(at)cisco(dot)com> wrote:
> I see.  Perhaps earlier I alienated some people by talking about
> transactions per second, which is not necessarily an appropriate unit.

Hard to say, tps is still important. A lot of loads need to change
one thing in a transaction and change it fairly fast. A lot of other
loads update / insert 10,000 rows, and tps doesn't mean as much as
rows inserted / second. I doubt anybody took it personally though.
More likely just got confused over what you and they were measuring.

Storytime!

One of my best friends came from MSSQL and MySQL shop, and when
introduced him to pgsql, he was bitching at how slow this file full of
inserts was inserting. This was in the pg 7.0 days, when pg was not
super fast, but quite fast enough if you knew a few basic tricks of
the trade. I looked at his load file, which was basically one insert
after another, 10k of them. I pushed him aside, typed in begin; on
top and commit; on the bottom and told him to run it now. It finished
in about 10 seconds. It had taken > 10 minutes before that. He's now
my boss at another company, and pretty hard core pgsql fan.