Test how bcache performs when using more than one HDD and a single SSD cache.
The experiment should be setup to illustrate a problem we expect bcache will have--if cache space is not intelligently divided, then a greedy VM that makes many requests will get more of it than a slower VM, even though the slower VM might get more benefit out of the space. The vcacheshare paper has a similar experiment.
You will need a way to control the access rate of the processes reading from each disk. You probably won't be able to just use a standard IO benchmark since they all try to read or write as fast as possible. I would focus on the reading case since if you use a write through cache the SSD won't speed up writes anyway.
Test how bcache performs when using more than one HDD and a single SSD cache.
The experiment should be setup to illustrate a problem we expect bcache will have--if cache space is not intelligently divided, then a greedy VM that makes many requests will get more of it than a slower VM, even though the slower VM might get more benefit out of the space. The vcacheshare paper has a similar experiment.
You will need a way to control the access rate of the processes reading from each disk. You probably won't be able to just use a standard IO benchmark since they all try to read or write as fast as possible. I would focus on the reading case since if you use a write through cache the SSD won't speed up writes anyway.