I have a server we just installed with Raid 10 4 disks at 7200 RPM sata drives. When copying a large file say 150GB it took about 5 hours. The raid is dell hardware perc controller, and the server specs are dual quad core 3.0 GHZ. Using nmon we can see the disks spiked out and the wait time.
My question is, what would you say the normal time to copy 150GB on the same HDD just a standard cp . to .new should take on this Raid10 config? I was pointed in the direction of perhaps the journaling of ext3 may be affecting this, what are your guys thoughts on this? Thanks in advance for your thoughts. Also when using mysql to crawl data on a very large database, which is the reason for this server, it crawls to a halt and becomes almost inoperable.

Well. First off, all not hardware raid controllers are created equal. Do you know what model PERC you got? It may be a low-end PERC, which is likely considering you are using just SATA drives.

2nd, large i/o operations will always slow the system down. This is largely down to the I/O bandwidth rather than CPU speed. If you need that kind performance, get yourself a nice high-end PERC H700 and some 20K SAS drives.

ongte,

Thank you for the reply. I checked omreport, this is what I got from omreport storage controller:

Controllers
ID : 0
Status : Ok
Name : PERC 5/i Integrated
Slot ID : Embedded
State : Ready
Firmware Version : 5.2.2-0072
Minimum Required Firmware Version : Not Applicable
Driver Version : 00.00.04.17-4.31.z-RH1



I am getting prices for 15k drives, but we created a test server in house, raid10 7200 drives. Is there a more detailed way to get the PERC information detail? Thanks again.

PERC5/i is a decent RAID controller, but it's quite a few year old. Your bottleneck is probably the drives.

You can probably see more details using the OMSA Web interface, assuming you installed it.

ongte, again thank you for your help.

I created a test server in house, 7200 drives, same as on the production server. 4 drives setup in RAID10 same as prod server. The performance is literally 10x faster. I do see some differences though I was hoping maybe you could help me understand.

When I run df -h I get:

Slow prod server:
Filesystem Size Used Avail Use% Mounted on
/dev/sda3 1.8T 568G 1.2T 34% /
/dev/sda1 996M 57M 888M 6% /boot
tmpfs 7.9G 0 7.9G 0% /dev/shm

In house faster server:
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/VolGroup00-LogVol00
442G 204G 216G 49% /
/dev/mapper/isw_haicahifd_Volume0p1
99M 31M 64M 33% /boot
tmpfs 3.8G 0 3.8G 0% /dev/shm

Even though the in house and prod server both have hardware raid.

also, when I run nmon to check disk performance while doing a large file copy benchmark, I see different disk configs:

Prod slow server:

³DiskName Busy Read WriteKB|0 |25 |50 |75 100| ³
³sda 0% 0.0 0.0| > | ³
³sda1 0% 0.0 0.0|> | ³
³sda2 0% 0.0 0.0|> | ³
³sda3 0% 0.0 0.0

Fast in house server shows:

│DiskName Busy Read WriteKB|0 |25 |50 |75 100| │
│sda 1% 0.0 394.4|W |> │
│sdb 1% 0.0 394.4|W |> │
│sdc 0% 0.0 382.5|W |> │
│sdd 0% 0.0 382.5|W |> │
│dm-0 0% 0.0 394.4|W |> │
│dm-1 0% 0.0 382.5|W |> │
│dm-2 1% 0.0 776.9|W |> │
│dm-3 0% 0.0 0.0|> | │
│dm-4 0% 0.0 776.9|W |> │
│dm-5 1% 0.0 776.9|W |> │
│dm-6 0% 0.0 0.0| |>

I know on my in house server each sata drive is connected to a port on the MB. I am sure with the PERC controller it must be different, although I dont have access to the server physically.

With that in mind, I assume since the drives are all 7200 SATA both in house and prod, it must be something else? What are your thoughts on this?

I will try and install the interface to get more details as well.

Thank you.