Subject: Re: dinode: di_spare[2]
To: Jonathan Stone <jonathan@dsg.stanford.edu>
From: Jason Thorpe <thorpej@nas.nasa.gov>
List: tech-kern
Date: 12/01/1997 19:02:16
On Mon, 1 Dec 1997 16:36:22 -0800 (PST)
Jonathan Stone <jonathan@DSG.Stanford.EDU> wrote:
> NAS=Numerical Aerospace Simulation (or something close)
>
> A NASA group at NASA Ames, (at Moffet Field, just down the peninsula)
> doing aerodynamic simulation, virtual wind-tunnels, and other neat and
> data/CPU-intensive stuff. Several NetBSD developers and a core group
> member are employed there. I'm sure they'll correct the acronym.
Yup! :-) NetBSD developers who are also NAS staff do things ranging
from network r&d, mass storage development, parallel systems development,
and graphics systems development.
> HSM==Hierarchical Storage Management
>
> Mainframe-style on-demand automigration of data between disks and
> less-accesible media, like tape robots. I think NAS still use this;
> they used to have a custom-built tape robot in the 80s, when their
> online storage demands exceeded the capacit of COTS solutions.
Actually, the tape robots are "COTS" ... they're large StorageTek silos
w/ Timberline-series drives. The original version of NAStore (our HSM
system) ran on an Amdahl. That was replaced several years ago with
a couple of Convex 3820s (which is what it still runs on). The software,
however, is not COTS. NAStore is made up of several userspace daemons
that do the data migration, a large database that maps file handle to
silo/tape/file, as well as modifications to the operating system's
UFS implementation, which provide the hooks for the data migration.
We're working on the next version, which will be cheaper, more reliable,
more scalable, etc., etc. Thusfar, most of that work is being done with
NetBSD/alpha on a few AlphaServer 8200s. We hope to possibly replace the
custom hooks with a freely-distributable DMAPI implementation (a standardized
API for HSM-style data migration) which will then be integrated into the
master NetBSD sources.
We still use HSM because of the insane amount of data we need to store
(mostly CFD data sets). We're projecting MSS 3 to be capable of storing
multiple petabytes in the first production version.
Jason R. Thorpe thorpej@nas.nasa.gov
NASA Ames Research Center Home: +1 408 866 1912
NAS: M/S 258-6 Work: +1 650 604 0935
Moffett Field, CA 94035 Pager: +1 415 428 6939