[dm-devel] dm-cache vs enhanceio testing

[Graeme H]

Hi,

This is a quick response to *msnitzer's *comments here:

http://lwn.net/Articles/541975/



A big workload for us that I'd like to see evaluated between all potential
kernel caching solutions in the sysbench OLTP benchmark against Percona
Server or MariaDB. It is directly comparable to many of the workloads we've
got in production, and I imagine is very similar to the workloads of
numerous other shops.

For simulation purposes, it's usually sufficient to set the InnoDB buffer
pool size to some fraction of the working set size, which will force the DB
to bring work in and out of RAM. That'll allow you to conduct a repeatable
test regardless of the size of physical RAM in the system.

The other use case that I'd like to see evaluated is hosting modern,
managed filesystems (such as btrfs and zfs) on top of dm cache devices. In
the past, this was quite difficult with FlashCache, and so it'd be useful
to have examples of this in the dm cache docs, especially as it relates to
boot issues.

Thanks!
Graeme Humphries
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/dm-devel/attachments/20130421/2628eaf6/attachment.htm>