eric%cirr.com@localhost (Eric Schnoebelen) writes: > [on dom0] > 65536000 bytes transferred in 1.338 secs (48980568 bytes/sec) > [on a domU] > 65536000 bytes transferred in 5.713 secs (11471381 bytes/sec) My experience is much older. On a 2006-era system with regular SATA drives (in RAID-1, raidframe), I got about 80 MB/s from the dom0 raw disk. I got a bit under 10% less from the dom0 filesystem (on a large file which is used to store the domU's disk). I got a bit under 10% less again when in the domU reading from the raw device. So when you experiment, measure dom0 raw dom0 filesystem, reading from the file used to back the domU (make it non-sparse!) domU raw domU filesystem, reading from a big non-sparse file Overall the NetBSD/xen slowdowns seemed entirely reasonable. But, the total rates were slower, so there may be another mechanism at play. Measuring the dom0 file and the domU raw vdisk will give you a more isolated view of xen xbd overhead.
Attachment:
pgpVOC4br8bRg.pgp
Description: PGP signature