Topic: Software raid5 performance sucks ?

[barn]

This is a spinoff of my previous thread: http://forums.contribs.org/index.php?topic=37182.0

<short recap>
When I copy from my xp to sme datadisk a software raid5, the network utilization goes in a distinct sawtooth pattern swingning between 25 and 5 % with a frequency of about 10 seconds. (on a 54mbit wireless)

When I unmounts my raid5 from /home/e-smith/files/ibays/Primary/
so my ibays now is on my single disk OS hd, I then have a consistent high network utilization of 27%.
</short recap>

So obviously something to do with my raid5

smeserver: dual celeron 500mhz, 256mb ram on a abit bp6 motherboard (widely known NOT to be the server mb of all times) with 4 ata ports, 2xata33 and 2xata66.
Only 1 hd per port.

1 hd for OS on the ata33, 10GB
3 in a RAID5 for data, one hd on ata33, two hd on ata66 all 300GB

I have the following questions for you experienced people

1) Should this hardware not be enough to keep up with at 54mbit network datastream ?

2) Can you see any obvious mistakes I have made ?

3) What information would you like to see to help determine the bottleneck, besides the bonnie result ?

4) Would the sawtooth pattern indicate a buffer not working somewhere?

5) The bonnie result below, is that "normal" or expected for mine setup? I have no idea wheter the values are high or low

Version  1.03       ------Sequential Output------ --Sequential Input- --Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
barn           500M  7009  85 24178  38 11613  22  8796  88 26305  28 172.8   3
                    ------Sequential Create------ --------Random Create--------
                    -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16   365  98 +++++ +++ 16365  99   371  98 +++++ +++  1290  96
barn,500M,7009,85,24178,38,11613,22,8796,88,26305,28,172.8,3,16,365,98,+++++,+++,16365,99,371,98,+++++,+++,1290,96

[Confucius]

one hd on ata33, two hd on ata66

That's for certain not the best option. Most techies know that combining different bus-speeds will only slow the fastest element(s) down. That's ofcourse when the combination can work, sometimes that even is a problem already.

Several times I've read (not fundamented) statements that using 3 disks in a RAID5 configuration isn't good either, should be more.

my 2 cents

[Stefano]

Only 1 hd per port.
1 hd for OS on the ata33,
3 in a RAID5 for data, one hd on ata33, two hd on ata66

as Confucius said, not the best option; I mean, this solution is definitely slow.

1) Should this hardware not be enough to keep up with at 54mbit network datastream ?

and keep in mind that:
- 54Mbit/s is theorical bandwidth.. we are talking about 6.75 MB/s.. in real world if you get 5 MB/s you are really lucky
- 54Mbit/s is total bandwidth you have.. in my personal experience, 3 or more wi-fi connection at the same time (3 or more pc acting on lan) is a "shoot on your foot" situation
- wireless connection is unstable

if your first interest is to serve files really quickly you should:
- get a sata controller and sata disks: keep in mind that pci bus is limited to 133 MB/s (theorical)
- get a wired, 1Gbit/s ethernet adapter

just my 2c

Ciao

Stefano

[barn]

I updated my machine with the latest patches (including some samba patches) and now the problem is solved...


To Neonano, I'm not really looking for a fast fileserver, I was interested in knowing why my fileserver had a performance that was much worse than you could expect of that hardware.


Erik

[Confucius]

Conclusion : RAID5 is ok... user was the problem