I usually simulate the load of the whole stack, by benchmarking the
application directly using a tool such as jMeter. Loading MySQL specifically
with read/write patterns similar to those of the real application can be
useful, but quite hard to accomplish.
One easy way of doing that is to enable
That only works if the application is already in production and gets a
certain load. For a new application (for example the database is
designed but the application is not ready yet) this doesn't work.
Amr Mostafa wrote:
I usually simulate the load of the whole stack, by benchmarking the
Anyone else? You guys don't simulate realistic workload in benchmark
just do sysbench or something like that? If there were a tool for that
(which can handle data generation, initial database generation, and
query generation, maybe it's output will be a jmeter test case) would
you do so?
I find it difficult to use a general tool to generate a lot of data specifi
to my application. There are many tools out there to execute queries against
a MySQL server it seems and log/report the performance. However, finding a
tool that is able to generate a lot of data that fits into your schema
Hi,
We have lots of mysql servers, master-slave and sharded databases. A
recurring task when a new feature/application comes in to test the
database with real workload. This needs test data and test query
generation. Until now I did this with ad-hoc scripts, I looked for tools
to do this, so
I use jmeter too, but it can't generate the test dataset (if I have to
write this, I plan that it will create a jmeter test case with the
generated test data). Usually my ad-hoc script generates csv files (to
load initial data) and jmeter test cases.
Michael Dykman wrote:
On Wed, Jun 3, 2009