On Thu, Sep 23, 2004 at 10:01:43AM -0500, Ivan Petrovich wrote:
> Are you suggesting that once Amanda has run once (for 8 hrs, using
> gzip both in "estimate" and "dump"), it should not need to run gzip
> for the estimate phase ever again,

My understanding is that Amanda *never* runs gzip in the estimate
phase.  For the first run, it guesstimates 50% compression, same
as the tape-drive marketing-bumf writers :-)  (Actually, Amanda
uses the value of the "comprate" amanda.conf parameter, whose
default value is 50%).

> My claim was based on experience of running a test configuration once,
> and observing an 8-hr run time, then after turning off software
> compression, observing that the time was reduced in half. I
> immediately abandoned software compression.

I'd expect software compression to slow down the run, but I'd
expect the slowdown to occur in the dump phase.  Are you sure
that the extra four hours you saw were spent in estimation?  If
so, I don't know how to explain your observations.

The amount of resulting slowdown would depend on too many factors
to be easily predictable, but if it came anywhere near doubling
the run time, I'd look for ways to tune it -- change from "best"
to "fast", move the compression between clients and server, etc.
(We compress most of our DLEs on the clients, but a couple of
ancient boxes are too slow for that, so their DLEs get compressed
server-side.)

--

|  | /\
|-_|/  >   Eric Siegerman, Toronto, Ont.        erics AT telepres DOT com
|  |  /
The animal that coils in a circle is the serpent; that's why so
many cults and myths of the serpent exist, because it's hard to
represent the return of the sun by the coiling of a hippopotamus.
        - Umberto Eco, "Foucault's Pendulum"