So does this mean that the architects scale down both the workload and the simulator machine configuration?
ChandlerBing
I think the second half of the slide means they just test the scaled down version in complete detail and be confident that it is a good indicator of the reality.
Corian
Even with an accurate simulator, how do testers predict accuracy on large data sets from smaller ones? I think I remember from lecture that data sets too large for smaller machines can result in super linear speed up. Can that really be predicted from results on smaller sets?
So does this mean that the architects scale down both the workload and the simulator machine configuration?
I think the second half of the slide means they just test the scaled down version in complete detail and be confident that it is a good indicator of the reality.
Even with an accurate simulator, how do testers predict accuracy on large data sets from smaller ones? I think I remember from lecture that data sets too large for smaller machines can result in super linear speed up. Can that really be predicted from results on smaller sets?