On Wed, Feb 03, 2010 at 11:32:43AM -0800, Matthew Dillon wrote: > > What I will be doing this week and possibly into next week is > implementing a system-wide cache-to-swap feature for objects backed > by the VM system. Such objects include file data and filesystem > meta-data. The type of data/meta-data to cache will be selectable with > a sysctl and the write-bandwidth to the SSD will also be manageable.
That would be awesome :-) Would it be possible to implement this as a multi-stage caching system ? Hard drives come in 3 or 4 different speed grades these days and a caching layer going from fastest drives to slow drives would make a lot of sense. We could have this situation in a file server: 32 GB SSD > 300GB in 15,000 RPM drives > 10s of TB in 7200 RPM drives Which is not too different from the multi-level caches of CPUs these days... -- Francois Tigeot
