I'm currently doing research on heterogeneous clusters. This is where you take machines with different processor speeds, network speeds, architectures, etc, and hook them up to run as a cluster. This is opposed to the standard cluster which is homogeneous with all machines the same.
We are looking into ways to evaluate performance of these clusters and the problem is that most of the computers we have running in the cluster right now are the same speed. I want to find a way to simulate lower clock speeds on these machines to in turn simulate a more heterogeneous cluster.
I used to be fairly into overclocking, so I know the ins and outs of changing clock speeds. I checked the bios on the machines and they're dells so they don't allow the changing of clock speed from within the bios.
I've looked at tools that allow you to change the FSB from within windows (these are pentiums so no multiplier adjusting possible), but our lab machines are all running on linux. It would be possible to switch ALL machines over to windows for testing, but it's generally frowned upon to do scientific work on windows
Any help? Thanks in advance.
We are looking into ways to evaluate performance of these clusters and the problem is that most of the computers we have running in the cluster right now are the same speed. I want to find a way to simulate lower clock speeds on these machines to in turn simulate a more heterogeneous cluster.
I used to be fairly into overclocking, so I know the ins and outs of changing clock speeds. I checked the bios on the machines and they're dells so they don't allow the changing of clock speed from within the bios.
I've looked at tools that allow you to change the FSB from within windows (these are pentiums so no multiplier adjusting possible), but our lab machines are all running on linux. It would be possible to switch ALL machines over to windows for testing, but it's generally frowned upon to do scientific work on windows
Any help? Thanks in advance.