[TUHS] NFS at 40
Matt Day via TUHS
tuhs at tuhs.org
Sat Sep 27 13:43:36 AEST 2025
You're right, according to Russel Sandberg's 1986 paper,
"The Sun Network Filesystem: Design, Implementation and Experience":
"The System V.2 port [of NFS] was done in a joint effort by Lachman
Associates and The Instruction Set on a VAX 750. In order to avoid
having to port the Berkeley networking code to the System V kernel
an Excelan board was used. The Excelan board handles the ethernet,
IP, and UDP layers. A new RPC transport layer had to be implemented
to interface to the Excelan board. Adding the vnode/VFS interface
to the System V kernel was the hardest part of the port."
-- https://nfs40.online/wp-content/uploads/2025/08/rusty-nfs-revised.pdf
A couple other paragraphs that caught my eye in Rusty's paper:
"The port to the IBM PC, done by Geoff Arnold and Kim Kinnear at
Sun, was complicated by the need to add a “redirector” layer to
MS/DOS to catch system calls and redirect them. An implementation
of UDP/IP also had to be added before RPC could be ported. The NFS
client side implementation is written in assembler and occupies
about 40K bytes of space. Currently, remote read operations are
faster than a local hard disk access but remote write operations
are slower. Over all, performance is about the same for remote and
local access."
"At the UniForum conference in February 1986, all of the completed
NFS ports were demonstrated. There were 16 different vendors and
five different operating systems all sharing files over an ethernet."
"There were many people throughout Sun who were involved in the NFS
development effort. Bob Lyon led the NFS group and helped with
protocol issues, Steve Kleiman implemented the filesystem interface
in the kernel from Bill Joy’s original design, Russel Sandberg
ported RPC to the kernel and implemented the NFS virtual filesystem,
Tom Lyon designed the protocol and provided far sighted inputs into
the overall design, David Goldberg worked on many user level programs,
Paul Weiss implemented the Yellow Pages, and Dan Walsh is the one
to thank for the performance of NFS. The NFS consulting group,
headed by Steve Isaac, has done an amazing job of getting NFS out
to the world."
On Thu, Sep 25, 2025 at 6:39 PM Tom Lyon via TUHS <tuhs at tuhs.org> wrote:
> Lachman provided the reference NFS ports for System V folks, and lots of
> porting work for various vendors. IIRC.
>
> On Thu, Sep 25, 2025 at 5:36 PM Larry McVoy via TUHS <tuhs at tuhs.org>
> wrote:
>
> > Does anyone know why Ron Lachman was there? What did he do for NFS?
> >
> > On Wed, Sep 24, 2025 at 09:38:55AM -0700, Al Kossow via TUHS wrote:
> > > https://nfs40.online/
> > >
> > > this is all pretty cool
> >
> > --
> > ---
> > Larry McVoy Retired to fishing
> > http://www.mcvoy.com/lm/boat
> >
>
More information about the TUHS
mailing list