LinuxQuestions.org - Software RAID1 perfoming bad (expecting too much?)

- Linux - General (https://www.linuxquestions.org/questions/linux-general-1/)

- - Software RAID1 perfoming bad (expecting too much?) (https://www.linuxquestions.org/questions/linux-general-1/software-raid1-perfoming-bad-expecting-too-much-303941/)

Software RAID1 perfoming bad (expecting too much?)

Hello everybody,

In a firm where I work there is a need for file server that has to offer to its clients (Windows XP,2k) a bunch of files (various sizes), but all together around 300-500MB at once. So obviously I choose Linux to do the job and since I have read that SW RAID is equally (or better) capable as HW RAID I choose SW. Due to the need of read speed (write speed is not so important) and redundancy went with RAID1.

I've setup a computer using Debian GNU/Linux and while you read this it already does its job, the only thing that I'm bothered is that I can't get it to perform as I thought it would (it should?). I'm in doubt now: have I misconfigured something or have I expected too much from SW RAID 1 on such hardware?

I ask for your expertise.

Best regards, Blaz.

P.S.: I hope I have provided you with all the information needed. Should you need anything else, just say so. ()

FILE SERVER:
*ASUS CUSL2, Intel 815E, ICH2, ATA100(latest BIOS-1009)
*P3 800MHz (Socket 370)
*512MB PC133 SDRAM
*2xMaxtor 200GB, 7200rpm, 8MB cache, ATA133 (6B200P0)
*2x40-pin (80-conductor) cable
*Intel PRO-1000 MT Desktop Adapter
*Debian GNU/Linux Sarge

**********************************************************
* PLEASE CHECK THIS! >>> htp://loophole.mine.nu/raid_perf/links.htm *
**********************************************************

CLIENT 1:
*AMD Athlon 64 3200+ (2.0GHz)
*1GB RAM
*Intel PRO-1000 GT Desktop Adapter
*Hard disk: Maxtor SATA
*Microsoft Windows XP SP2

CLIENT 2:
*Intel P4 2.8GHz HT
*1GB RAM
*Intel PRO-1000 MT Desktop Adapter
*Hard disk: IBM (I think they write max ~25MB/s)
*Microsoft Windows XP SP2

CLIENT 3:
*Intel P4 2.6GHz HT
*1GB RAM
*Intel PRO-1000 MT Desktop Adapter
*Hard disk: IBM (I think they write max ~25MB/s)
*Microsoft Windows XP SP2

OTHER EQUIPMENT:
*3com OfficeConnect Gigabit switch 8

-----

iptraf measurements:
*testfile.zip (823MB) over Samba:
-FS > C1: ~22MB/s
-FS > C1, C2: ~23MB/s (jumped to 13MB/s after 10 seconds of transfer and then up again...)
-FS > C1, C2, C3: kept falling from 25MB/s to ~3MB/s at the end
-C1 (I used Linux; just for comparison) > C2, C3: 55MB/s

Well you really should isolate the problem first.

If you think it is at the raid level you should test the disk throughput directly, not via some other layer.

hdparm -t /dev/drive

I'm not sure how it will work on a software raid device, but try it out anyway.

That's where you want to start trouble shooting.

Then of course, test FTP transfers before checking samba.

You need to isolate where the problem is rather then just assume "My drive is slow because samba doesn't transfer fast."

What filesystem type are you using on the server?
ext3
reiserfs
xfs
etc....

Do you have any numbers from hdparm on the Debian machine?
How about network throughput/latency?
Redundancy is always welcome, but software RAID does have overhead.
With a reliable backup system, tape, dvd, etc., you might consider LVM on the server with an xfs filesystem.

Haven't done any benchmarking on my LAN as of late but my hardware doesn't
nearly match yours. I'm still at 100Mbps over Cat6.

nz

Quote:

Write-caching issues : many modern drives support write-caching. Sometimes, enabling it may be a huge performance win, especially when you have many concurrent disk writes like what Qmail does. The problem is that the ReiserFS journal has to be written synchronously for a proper replay at bootup time. At least, everything written to it should be here after a system shutdown or a crash. In fact, there is only a problem in case of a power failure. The write cache is in the disk itself, and if you got a kernel crash, data will still be written later. So, the following trick looks like a dirty hack but may help a lot with some drives. If you have an UPS, enable write caching by default, and configure your UPS daemon to automatically disable write caching when a power failure occurs.
FYI, this text is written on a Sony Vaio PCG-Z600RE laptop whose hard disk I/O is twice faster after the following command :
/sbin/hdparm -m 16 -p -W 1 -u 1 -c 3 -d 1 -X 66 /dev/hda

http://www.linuxfromscratch.org/blfs.../reiserfs.html

A little reading is in order here. cylix steered you to the hdparm tool and I asked what filesystem you were running. I now see you've got reiserfs. Some better low level benchmarks, and a little filesystem tuning are probably in order. Unless you expected everything to work right out of the can..... The top quote came from a googled site for 'reiserfs tuning'. and the URL next suggests you read 'man reiserfstune'. 'man hdparm' might also help. And, do not just fault your server, but, consider the clients too.

nz

nonzero:

I have provided hdparm -vi (http://loophole.mine.nu/raid_perf/hdparm_hda), where it can be seen, that I have write caching enabled (btw. this does not affect reading much; is good though).

I'm using: hdparm -d1 -c3 -m16 -u1 -W1 -X udma5 /dev/hd(a/c)

I have read reisesfstune and all that can be done with it is to move and change journal parameters. Journal is used when an application writes to disks, so again, shouldn't hit the read performance... (I think. Right?)

I have considered clients too, that is why I tested the disks on really fast machine (Client 1). I got 55MB/s there... much better, but still...

Does it run bad? I don't know, I just think it should go faster than this. I can't compare it as this is my first RAID box... I'd just like to know the answer to the thread topic... So anyone with comparable configurations, can you please write your numbers here?

And... of course I don't think everything should work out of the box, I have tried many configurations (hw&sw) before posting to LQ.

Thank you, Blaz.