Drop compression level to 1 or 2 (anything above 2/3 is generally a waste of resources and will require ever increasing amounts of cpu and time, with little/no gain in space savings).
Try to get your database and dump devices working across as many controllers and disks so as to minimize IO bottlenecks. Moving db devices will obviously take some planning; spreading out your dump devices is a simple change of the dump/stripe paths.
Make sure your dump devices are not mounted on a slow network/nfs mount point.
I'd also suggest you get your system and disk-subsystem admins involved in monitoring your dumps (and loads?) for bottlenecks (cpu? disk IO speeds? network IO speecs?).