Summary NAS question

• Previous message: Paul Reilly: "SRM & Alpha Bios?"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

Date: Fri, 30 May 2003 11:51:27 +0200
To: 'Tru64 User Group' <tru64-unix-managers@ornl.gov>

Hello Managers ,
                  Here is the summary of my question posted asking "
comments on NAS solutions for my Tru64 boxes .

Thanks to all who responded and here are the comments in chronological
order of the posts .

>Michael Wheelock>>>>>

We have used a NAS solution from netappliance (www.netapp.com). It has all
of the scalability and reliability that most people need. You can add disk
shelves on the fly, etc. Very good solution. Possibly pricey (though I
don't know your budget).

>Dan Goetzman>>>>>

Hello. I manage a 2 node TruCluster NFS (and ASU for CIFS/SMB) server that
us our largest and fastest file server on our network. Running 5.1A PK4, 2
ES40 nodes, two ESA12000 StorageWorks SAN attached storage (currently 3 TB).

  Many things to comment on, let me see...

  Very fast NFS server. NFS on Tru64 is kernel threaded, the Alpha
processors
are very fast (more that we need!) and we are gigabit connected to our
network backbone.

  Reliability is very good. We have been running 2 years and have yet to
have
a disk failure in the StorageWorks units. The hot spares are waiting to be
called into action. I cannot say enough good things about the StorageWorks
units. NFS has been tweaked a bit under TruCluster to survive a node failure
(like a panic) and the NFS clients will never know the difference. Nice! We
sometime do what I call a "rolling reboot" (reboot one node at a time) if we
feel the cluster needs it for some reason and the NFS clients just keep on
file serving to the cluster.

  Expansion? We purchased the StorageWorks units half populated. We can add
storage online with out a reboot any time we want. TruCluster makes adding
another node REAL easy.

  Pitfalls? We built our cluster NSPOF (no single point of failure).
Anything
less and you might as well run a single point of failure server, like say a
Sun box. We looked at other vendors and the TCPIP stack and NFS
compatibility was a big factor. Tru64 seems to have done a reasonable job
here. HP-UX and AIX are weak in this area. Of course, what ever Sun does is
"right".

  Also, I might add that there are other things that need to be taken into
account. The DNS server going down will stop things like mountd from
responding and make your super bullet proof NFS server appear to be down.
Slow DNS server response, the NFS server will appear slow.

  Also, TruCluster is probably the easiest cluster solution to support. But,
you will need to get someone trained on all the in's and out's of TruCluster
and it will have problems (the patch just like everyone else). A GOLD
support contract would be nice if you do not have strong Tru64 skills. Our
site has about every NFS client (and OS version) that has ever been made, so
we seem to always have a few NFS related "Customer Special Patches".
Although that list is very short when I moved to 5.1A.

  Having said all this, I must say our TruCluster NFS server is still the
big, bad, fastest file server we own. The standard by which replacement file
servers are being measured. Doing it again today? That would be interesting.
We are looking at building a Linux cluster to do the same thing. Advantage?
Cost. Disadvantage? Linux clusters are not there yet when compared to
TruCluster. Also, we looked at using a Sun cluster, but Sun Cluster 2.x was
not very good. Their 3.x product gained much lost ground to TruCluster so we
would have to look at it. Although I have not had any hands on the new Sun
Cluster product.

>Paul MAGLINGER>>>>>

We've been running a Network Appliance F740 for close to 4 years. Best
investment we've ever made. It runs and runs and runs. Adding drives to
volumes is a breeze, performance is great, and our DBAs love it.

• Previous message: Paul Reilly: "SRM & Alpha Bios?"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]