Tests bind EF in a different lifecycle to the other frameworks

[lukemcgregor]

I'm always a bit skeptical when I see differences of more than a factor 2 when looking at benchmarks. In reality most frameworks are pretty close in performance with the exception of edge cases around specific functionality.

I think that your specific 3-6x slower figure is coming from the way you manage your lifecycle.

https://github.com/ppanyukov/EntityFrameworkBenchmarks/blob/master/src/EntityFrameworkBenchmarks/Startup.cs#L22-L29

You are using a completely different type of lifecycle (transient vs singleton) for EF and dapper. This means (amongst other things) you are creating a new connection when querying EF and reusing the connection with dapper.

[ppanyukov]

I will check it out. Even if that's true, the point stands (somewhat) in that out of the box other frameworks and libraries perform better out of the box without understanding all these complexities such as lifecycle etc.

[lukemcgregor]

So your argument is that because you don't understand a framework its bad? Sounds a little flawed to me.

[ppanyukov]

Not quite. I guess the original benchmark intent was "test default behaviours" if I can remember correctly. Obviously lack of knowledge of the tool is not the tool's fault, at the same time the defaults of other frameworks result in a much better results :)

Sent from my iPhone

On 29 Feb 2016, at 19:39, Luke McGregor notifications@github.com wrote:

So your argument is that because you don't understand a framework its bad? Sounds a little flawed to me.

—
Reply to this email directly or view it on GitHub.