Erik Hjertén <erik.hjerten AT companion DOT se> wrote on 05/24/2013 12:15:22 PM:

> Hi all
>
> I have invested in a used HP Proliant ML150 G5 server as a new
> backup server. I have about 500 GB of data in 40 000 files spread
> over 8 clients to backup. Data doesn't grow fast so I'm aiming at
> two 1TB disks in a raid 1 configuration.
>
> Do I go with more expensive, but faster (and more reliable?), SAS-
> disks. Or is cheaper, but slower, S-ATA disks sufficient? I'm
> guessing that disk speed will be the bottle neck in performance?
>
> Your thoughts on this would be appreciated.

What is your backup window?  12 hours?  You could do that with a *single* 7200 RPM SATA drive.  8 hours?  Probably still, but you'd have to do some testing to see.  Less than that?  You're going to need to intimately undertand both your circumstances and the various technologies inside of BackupPC to be able to answer that better.

Frankly, a mirrored array isn't gonna buy you much performance increase.  It won't help write performance at *all*, and I'm not sure you'll need enough read performance to matter:  the high amount of seeking that BackupPC requires doesn't really hep for getting sustained high read performance.

I will take Les' advice (don't use the Green drives) one step farther:  I recommend the drives designed for DVR/Video use.  "Normal" drives (the not-Green drives) are warrantied only for 8x5 usage;  the DVR drives are rated for 24x7 usage.  They're a little more expensive, but not much.

There are other questions you will need to ask that will make as much (if not more) difference than the speed of the drives you'll be using:

* Would more drives (even if they're slower) give you better performance?

* How fast can the *clients* push the data?  If you're limited by them, improving the server won't help?

* What is the speed of the network involved?  Are you talking 100Mb/s or slower?  That will severely limit your performance.  Do you Gigabit everywhere between them?  Are there points in between that might cause problems (like if the clients and server are in *different* switches)?  Can you do bonded Ethernet on the server?

* What technique are you using to back up the files?  Rsync over ssh (with encryption overhead)?  Rsyncd?  tar/SMB (which are much less intelligent in transferring files, but maybe less disk-intensive)?  Will you use compression on the server, and what level of CPU do you have?

* What do your files look like?  40,000 files for 500GB of data is a pretty high size-per-file.  (Contrast with one of my servers, which is 800GB, but 400,000 files: 1/5 the data per file.)  Are your files mostly small (say, under 10kB), mostly average (10k to 10M)?  Do you have any massive files (1GB or larger) to deal with?  Backing up a database server and backing up a mail server require *noticeably* different approaches.

(Believe it or not, even all of *this* is not all that's involved!)


Your question is along the lines of "Which is faster, a bicycle or a dump truck?"  The answer is:  it depends.  Need to move a mountain of sand 50 miles away?  A dump truck.  Delivering urgent letters in Manhattan?  Bicycle!  :)  I know you tried to give us some idea of what you're trying to do (500GB and 40,000 files across 8 clients), but not nearly enough to accurately answer the question.

But keep this in mind:  the numbers that you supplied are what I would consider small and "boring":  a relatively small amount of data and a relatively small number of clients.  The only thing I find interesting is that your 500GB of data is only 40,000 files.  That's a small number of files, so I am curios as to the size of your files.  Other than that small eybrow-raising item, your application is extremely straightforward--assuming that the answer to all of the questions above are the typical answers I would expect in a decent office environment.

Tim Massey

Out of the Box Solutions, Inc. Creative IT Solutions Made Simple! http://www.OutOfTheBoxSolutions.com tmassey AT obscorp DOT com		22108 Harper Ave. St. Clair Shores, MI 48080 Office: (800)750-4OBS (4627) Cell: (586)945-8796

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service 
that delivers powerful full stack analytics. Optimize and monitor your
browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may

_______________________________________________
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/

--

---

Erik Hjertén erik.hjerten AT companion DOT se 0708-90 55 30 Näsvägen 40, 139 33 Värmdö 08-403 750 50 www.companion.se