|
Re: [BackupPC-users] Centralized storage with multiple hard drives
2014-03-20 16:09:58
Les Mikesell <lesmikesell AT gmail DOT com> wrote on 03/20/2014
03:48:51 PM:
> It's all about statistics and the odds of having to move the disk
head
> to get a directory entry or inode vs already having it in cache for
> instant access (and sometimes even some data...).
In theory, theory and practice are the same. In
practice, they aren't. (Or, another way: theoretical analysis
doesn't always make a difference in the real world.)
You keep telling me the theory of why more RAM would
help. I keep telling you the FACT that it doesn't. Repeating
your theory doesn't change the FACT that it doesn't help.
(Why don't you try it sometime? Either use a
kernel parameter or remove some RAM from one of your magical 20GB boxes.
You stated that even 4GB was not enough to see the effect you're
describing, so do that. I *HAVE* done this test. Several times.
You only have to do it for a single full backup of a single large
target: one night and you'll have a FACT to analyze, not a theory.)
> I care because _some_ of my targets are relatively
fast - and have new
> big files daily. These will routinely show a 40+ MB speed in
the
> backuppc host summary, although I'm not exactly sure what that is
> measuring. 10 or 12 is more common for run-of-the-mill targets,
even
> less if the files are mostly small or the server is older and slower.
If you don't know what it's measuring, how can you
use it for anything.
> > And my point was not that I *am* limited by my 1GbE link (I'm
not
> usually with BackupPC but I *am* with other uses of my backup
> servers in general),
>
> It seemed like you were advising to expect backuppc to be limited
by
> bandwidth - which doesn't match my experience at all once you get
the
> initial copy.
Nope. Merely (and somewhat awkwardly) making the point that it is
*meaningless* to make your disk performance faster (as in huge RAM caches
and avoiding parity RAID) if it's already several times faster than some
other bottleneck.
The idea encouraging someone to not using RAID-5 or
-6 on a BackupPC box is just about the biggest waste of money I can possibly
imagine for such a machine. Well, maybe encouraging someone to use
15,000 RPM SAS drives not in a parity array might be worse... :).
The cost of hard drives *always* exceeds 50% of my array-based (as
in not single drive) BackupPC servers. Why would I want to make that
storage even *more* expensive if it buys me no measurable difference in
performance?
And make no mistake: I have not personally seen
an application where lack of RAM (where that RAM is measured in Gigabytes)
or disk performance (with 4 or more drives, let's say) has led to a bottleneck
on my systems. Now, I've not tried backing up more than low double
digits of targets to a single server, and I've not done much more than
10TB of pool space: maybe if you scale significantly beyond that,
maybe. But even then I'd need to see it, not just theorize it...
And I'm officially done! :) (Unless there
are hard facts presented.)
Tim Massey
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech _______________________________________________
BackupPC-users mailing list
BackupPC-users AT lists.sourceforge DOT net
List: https://lists.sourceforge.net/lists/listinfo/backuppc-users
Wiki: http://backuppc.wiki.sourceforge.net
Project: http://backuppc.sourceforge.net/
|
|
|
