v9fs-2.1
The
principal
problems
seen
during
v9fs-2.0
use
is
the
difficulty
in
the
setup
of
the
server
and
performance
problems
associated
with
multiple
extraneous
copies
in
standard
data
flow.
This
is
particularly
the
case
when
trying
to
optimize
communication
using
intelligent
networks
or
shared
memory
between
virtualized
partitions.
As you can see from the above illustration, each white arrow involves a copy. Also, if you want to access a remote synthetic file system you end up traversing the v9fs layer multiple times.
In version 2.1 we will try to address these problems by eliminating unnecessary copies from the transport path and moving portions of the server into the v9fs kernel code.
In the 2.1 design we move the servers into the client allowing short-circuiting local access to synthetic file systems and resources including synthetic gateways to system devices. The main point is to help prevent multiple copies through user-space when exporting local devices (and synthetic gateways to those devices) through the server.