On Fri, 9 Feb 2001, Shaw Terwilliger wrote:
> I've been using PostgreSQL for a few years now in various projects, and
> I've been very happy with its features and performance.
>
> I have a client who wants to know how PostgreSQL (7) will perform over
> NFS. The NFS client is Solaris, the server is a big network storage
> appliance. The network is probably gigabit ethernet.
>
> My first reaction is to tell them they're just crazy, and that's not how
> a database is intended to work, and they should buy some fast, redundant
> local storage. I thought I'd ask the designers and users here so I could
> back up my recommendations.
we use mysql over nfs here. it works fine. the nfs server is a netapp.
it's mounted over a gigabit ethernet crossover cable to the linux machine
that runs mysqld. it's unlikely that gigabit ethernet would be much of a
bottleneck on any modern disks, plus netapp's wafl filesystem is basically
faster than anything out there right now, so the only overhead is the
tcp/ip (udp) layer, which probably isn't that expensive. running it over
a 100megabit line would probably slow things down some, but it all depends
on how hard you're going to be pushing it.
so the issue left is reliability. i've looked over some of the replies
already that talk about unreliability in general, async issues, and
locking issues. here's what you do to avoid all of that: use an nfs
implementation that works. we use linux 2.2.18 with mount and nfs-utils
packages from http://nfs.sourceforge.net/. we use nfsv3, not v2, and we
mount the filesystem with the sync option. problem solved.
i haven't tried this with postgresql, but we actually are planning on
switching to postgresql down the road, so hopefully it'll be ok. heh.
-tcl.