DR Speed Question

[SHORT CUTS]

I was using Corona DR for the first time and have a quick question about the speed gain when using DR.

Here is the setup & test:
Workstation-Aurora rendered the scene in 10:33 min
Workstation-Zeus rendered the same scene in 20:25 min

Using DR, Aurora as the master and Zeus as the slave, the scene rendered in 10:02 min while Zeus was only contributing 7 passes in those 10 minutes.
There is quite a difference when it comes to rendering power between those two machines but i was expecting a bit more gain out of it. Saving 31 seconds wont justify powering up another workstation.
So i did another rendering on Zeus and checked how many passes the single rendering was able to do within 10 minutes -> it got to 14 passes which is double from what i got in the same timeframe when using DR.
Since setting up DR is pretty much fail save i dont think i did something wrong on this end, so my question is, is that expected behavior? (Maybe because of the power difference?)  I know that the 2nd workstation will start a bit later with the rendering but getting only half of the passes in the same amount of time when using DR seems off to me....

[TomG]

Hard to say, as it could depend on the complexity of the scene as to how long it takes to send it over the network for the second machine to have all it needs to begin rendering. At most you could expect around a 3 minute speed up, then it only need take a couple of minutes to send the scene for that to be reduced into the 30 to 60s range as the benefit.

You could try rendering the same scene for twice as long, see if the benefit is more significant then. You could try placing assets on a network drive, e.g. all the textures, so that they don't need to be sent over the network but instead both machines would read them from the same place (maybe for things like proxies too, maybe a pre-calculated and saved cache so that both machines don't need to separately calculate the cache).

[SHORT CUTS]

Finally had some time to test it a bit more and the speed gain with larger scene and way longer render times is very little, it shouldnt be the network, i guess the old workstation is simply too slow to benefit from DR. The new workstation did 50 passes while the old one only 10 using DR.
I will use the old one for animations or stick to our prefered render farm. thanks for you help.

[muoto]

Have any studies been made with a given average complex scene, fast 1gb network (usual set-up i guess) and multiple identical render stations, to measure what the actual speed gains are ?

I guess that 2 same render nodes does not give you exactly x2 speed, it should probably be a bit less (time sending the scene via DR, starting, loading the scene etc, it takes some time), and 3 render nodes not exactly x3 etc..

Is there a bottleneck when having too many render nodes ? (network?)

How does it compare to vray's bucket rendering system ?