threads and VMS [Re: dce/rpc "client" api]

[Luke Kenneth Casson Leighton]

On Wed, 23 Aug 2000, Bob Mastors wrote:

> A single NT client running netbench will have multiple outstanding
> CIFS commands to a server at many points in time. In other words
> the NT client will not always wait for a CIFS command to complete
> before issuing another one.
> To take advantage of this you could keep the samba model of one
> smbd per connection but redesign smbd to use posix threads.
> This won't make the security context issues any different than
> in the current smbd design.

the other method is to use an implementation of the VMS "IO Queue"
structure, which chris frank of linuxcare suggested.

a shared-memory "io queue" in which a worker smbd PROCESS [not thread]
feeds SMB requests in at the back, and a series of smbd PROCESSES pick up
from the front, one at a time.

another shared-memory "io queue" feeds the SMB responses back.

the potential performance hit of copying in-and-out of shared memory could
be avoided by using the shared-memory buffers directly in read_data and
write_data calls.

what do people think?

[thx bob].