Just curious about performance

Up until about 2 or 3 weeks ago I ran my server at home (with FreeBSD) for 3 years on an HP dc7100. The specs of this machine were ancient as one would expect for an 8 or so year old desktop:

Pentium 4 2.8Ghz
1GB RAM
Dual 80GB SATA drives mirrored (gmirror)
UFS file system

But the machine has run amazingly well for what I use the server for.

A couple weeks ago I had the opportunity to get (for free) an HP dc5800 (which I migrated to):

Dual Core 2 3Ghz
4GB RAM
Dual 320GB SATA drives mirrored (ZFS)
ZFS file system (encrypted with GELI)

My old machine never felt slow...unless I was recompiling all my ports again or recomping my kernel. But the funniest thing is, the "new" machine I have moved onto doesn't feel much faster at all despite the hardware being much more modern.

Is this just because FreeBSD is so amazing on old hardware that putting it on something newer doesn't make much of a difference?

I'm just surprised thats all. Curious to know what other board members think or have experienced! :)
 
xy16644 said:
  • 4GB RAM
  • Dual 320GB SATA drives mirrored (ZFS)
  • ZFS file system (encrypted with GELI)

Could be your file system choices have something to do with how you perceive performance. Adding RAM may help; tuning may help; tuning with more RAM would be optimum.

I wonder how you'd perceive performance if you had a regular UFS file system, mirrored or not.

ZFS / mirrored (but not encrypted) on my fairly high spec i7 / 16GB RAM two fairly fast 1GB drives mirrored felt pretty quick. On UFS it doesn't feel appreciably faster, not for the work I do on it. I did not do any real benchmarking.
 
If it's a web server he might not notice any change cause the web can only move so fast, especially from a typical home connection. I still have two servers running P3s with 192K memory (talk about ancient) that still work great but I can't put anything more on them cause the memory would get swamped. But you don't need a lot of processing power to serve web pages unless there are a lot of hits or their dynamic.
 
Thanks for the replies! I should mention that I'm not unhappy at all with the "new" machine I have...I was just expecting a little more I guess. Having said that, ZFS with GELI is amazing and I'm really pleased I moved over to these technologies.

I was thinking about getting the Soekris hardware accelerator card to increase encryption performace but the shipping to where I live is almost as much as the card!
 
@xy16644

GELI is what's keeping the performance down in your case. Core2 lacks the AESNI hardware encryption that accellerates performance to over native speed. But as you said, you're not unhappy with it. I myself have a Core2 Duo laptop with ZFS/GELI, and even though the performance isn't that awesome, I'm happy just to have the encryption:)

/Sebulon
 
Luckily its a low traffic server ;-) Its only really busy when the backups run, I'm upgrading ports or doing an OS upgrade.

Is there a quick and easy way to check the throughput I am getting with ZFS/GELI?
 
@wblock

Well, yeah, but it tests an individual disk. In this case OP wants rather to know the average of his filesystem, which actually are two drives, geli'd with mirrored ZFS. That performance is waaay different from what a drive performs by itself.

# bonnie++ -u 0(if root) -d /foo/bar -s Xg(RAMx2 or more)

/Sebulon
 
True, but diskinfo(8) will let you see the speed of the underlying device. It will work on devices in /dev, including the GELI device, or GPT or filesystem labels, like /dev/gpt/usrfs.
 
Back
Top