Coda File System

Re: LWPs in Coda and native platform threads

From: Robert Watson <robert_at_cyrus.watson.org>
Date: Sun, 26 Jul 1998 18:12:06 -0400 (EDT)
On Sun, 26 Jul 1998, Shafeeq Sinnamohideen wrote:

> On Sun, 26 Jul 1998, Jim Doyle wrote:
> > What are the major feats of moving away from LWP's and to native
> > threads?  Ultimately, a proprietary user-space threads package 
> > is a liability IMHO - as we learned with DCE threads on FreeBSD.
> 
>   As Peter said, the main problem is that LWPs are by definition
> cooperative, not preemptive, and all of Coda is written with that
> assumption. Its main benefits are that thread switching is fast and
> doesn't involve the kernel, and it provides just enough functionality for
> what Coda needs. It would not be too hard to base LWP on top of, say,
> pthreads, or something similar. The main benifits that I can see would be
> portability and better I/O handling, especially wit a kernel ptherads
> implementation. 

Also, the benefits of parallelism on an MP machine -- with increasing need
for strong encryption (bye-bye to DES as an option as of last week), the
servers may become more CPU-intensive.  With true preemptive pthreads
support, SMP pthreads support (such as in Solaris, and presumably other
platforms such as Linux+*BSD) would be possible. 

This would be quite a bit of work, but the results would be quite good --
especially if in the future there are ports to Java, etc, where threads
are more common-place than present-day UNIX.

  Robert N Watson 

Carnegie Mellon University            http://www.cmu.edu/
TIS Labs at Network Associates, Inc.  http://www.tis.com/
SafePort Network Services             http://www.safeport.com/
robert@fledge.watson.org              http://www.watson.org/~robert/
Received on 1998-07-26 18:12:33