reiser is way faster than ext3

illtbagu · 12-17-2003, 10:54 PM

Wow what a differnece using reiserfs over ext3, reiser is almost 2x faster on my usb 2.0 hard drive. I can't believe it. Take a look at this

/dev/sda1 1 19929 160079661 5 Extended
/dev/sda5 1 9561 76798669+ ext3
/dev/sda6 9562 13640 32764536 fat32
/dev/sda7 13641 19929 50516361 reiserfs

/dev/sda7:
Timing buffered disk reads: 64 MB in 2.71 seconds = 23.62 MB/sec
[root@schrock321 schrock]# /sbin/hdparm -t /dev/sda6

/dev/sda6:
Timing buffered disk reads: 64 MB in 5.10 seconds = 12.55 MB/sec
[root@schrock321 schrock]#

Jestrik · 12-18-2003, 09:54 AM

Yes, i had read somewhere (micromart if you're from the UK) that it was faster... Didn't realise that much faster though

michaelk · 12-18-2003, 11:36 AM

Quote:

Originally posted by illtbagu

/dev/sda1 1 19929 160079661 5 Extended
/dev/sda5 1 9561 76798669+ ext3
/dev/sda6 9562 13640 32764536 fat32
/dev/sda7 13641 19929 50516361 reiserfs

/dev/sda7:
Timing buffered disk reads: 64 MB in 2.71 seconds = 23.62 MB/sec
[root@schrock321 schrock]# /sbin/hdparm -t /dev/sda6

/dev/sda6:
Timing buffered disk reads: 64 MB in 5.10 seconds = 12.55 MB/sec
[root@schrock321 schrock]#

I would hope that reiserfs is faster then a FAT32 partition in linux or is that just a typo and you really meant sda5.

From what I've read: reiserfs is faster with smaller files, jfs is faster with big files. ext3 is just an ext2 with a journal but you can still use all of the ext2 disk utilities if you remount it as ext2.

illtbagu · 12-18-2003, 02:13 PM

ext3 and fat32 always run about the same speed for me. This was not a typo just a messup. I meant to test sda5. I will test the ext3 partition when I get home. But I'm confident that the fat32 and ext3 will be the same. I have tested the ext3 partition 2 days ago against the fat32 partition and they were the same. But I will post the actual results just the same.

Does anyone else care to post some comparisons also?

Jestrik · 12-18-2003, 02:25 PM

How did you benchmark the drive?

illtbagu · 12-18-2003, 05:30 PM

I'm assuming you want to know how I came up with these results? If so then do this at the command prompt, substituting sda6 with what ever your hard drive is
/sbin/hdparm -t /dev/sda6

for details do a
"man hdparm"
at the command prompt

lostboy · 12-20-2003, 10:27 AM

You might find this interesting :

Timing buffered disk reads: 64 MB in 1.95 seconds = 32.82 MB/sec

This is with ext3. To be exact, here is fstab entry :

/dev/hda3 / ext3 defaults 1 1

The machine is a 1.7g P4, on an Asus P4b533 board. The disk drive is a Maxtor 7200 rpm, 80g, UDMA100.

What I find most interesting is that this number beats those obtained by your scsi drives. I am reading that correctly, right ?

JC

lostboy · 12-20-2003, 10:44 AM

This is also very interesting :

root@jac3:~# hdparm -t /dev/hda2

/dev/hda2:
Timing buffered disk reads: 64 MB in 2.29 seconds = 27.95 MB/sec
root@jac3:~# hdparm -t /dev/hda3

/dev/hda3:
Timing buffered disk reads: 64 MB in 2.15 seconds = 29.77 MB/sec
root@jac3:~#

This is a much more modest machine ( PII 400mhz ).
But it does have a 7200rpm, 80g, UDMA100 drive. Maybe that is where the performance is coming from.

But these scores are also higher than the ones listed by illtbagu.

illtbagu, are you sure that something isn't mis-configured on your box ? It seems to me that with SCSI devices, you should be able to beat all my scores.

Please post again and list the SCSI controller and disk that you have.

JC

dalek · 12-20-2003, 11:15 AM

I always use the command

Code:

hdparm -tT /dev/***

put you drive where the *** is.

Code:

[root@smoker windows]# hdparm -Tt /dev/hda

/dev/hda:
 Timing buffer-cache reads:   128 MB in  0.33 seconds =387.88 MB/sec
 Timing buffered disk reads:  64 MB in  2.59 seconds = 24.71 MB/sec
[root@smoker windows]# hdparm -Tt /dev/hdb

/dev/hdb:
 Timing buffer-cache reads:   128 MB in  0.32 seconds =400.00 MB/sec
 Timing buffered disk reads:  64 MB in 14.90 seconds =  4.30 MB/sec
[root@smoker windows]# hdparm -Tt /dev/hdd

/dev/hdd:
 Timing buffer-cache reads:   128 MB in  0.32 seconds =400.00 MB/sec
 Timing buffered disk reads:  64 MB in  9.85 seconds =  6.50 MB/sec
[root@smoker windows]# hdparm /dev/hda

/dev/hda:
 multcount    = 16 (on)
 IO_support   =  0 (default 16-bit)
 unmaskirq    =  0 (off)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 9729/255/63, sectors = 156301488, start = 0
[root@smoker windows]#

I need to update mine. These drives are capable of a bit more. These settings are pretty conservative. May post back the new ones.

Later

lostboy · 12-20-2003, 11:27 AM

Hey, that's cool. When you include 'T' you also test the cache.

Here's my results :

P4 1.7g machine :

root@jac:~# hdparm -tT /dev/hda3

/dev/hda3:
Timing buffer-cache reads: 128 MB in 0.37 seconds =345.95 MB/sec
Timing buffered disk reads: 64 MB in 1.94 seconds = 32.99 MB/sec

-------------------------------------------------------------------------------------------

PII 400m machine :

root@jac3:~# hdparm -tT /dev/hda3

/dev/hda3:
Timing buffer-cache reads: 128 MB in 1.01 seconds =126.73 MB/sec
Timing buffered disk reads: 64 MB in 2.15 seconds = 29.77 MB/sec

Thanks for the tip.

JC

dalek · 12-20-2003, 11:36 AM

Yep it is neat. Just have to be careful with those settings. Can really screw one up.

Here's mine after a bit of tweaking:

Code:

[root@smoker windows]# hdparm /dev/hda

/dev/hda:
 multcount    =  1 (on)
 IO_support   =  3 (32-bit w/sync)
 unmaskirq    =  1 (on)
 using_dma    =  1 (on)
 keepsettings =  1 (on)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 9729/255/63, sectors = 156301488, start = 0
[root@smoker windows]# hdparm -Tt /dev/hda

/dev/hda:
 Timing buffer-cache reads:   128 MB in  0.32 seconds =400.00 MB/sec
 Timing buffered disk reads:  64 MB in  2.36 seconds = 27.12 MB/sec
[root@smoker windows]#

It's a bit better.

Later

lostboy · 12-20-2003, 11:48 AM

I usually do this, and it's usually safe :

hdparm -d 1 -c 1 -k 1 /dev/hda

You should see output like this :

/dev/hda:
setting 32-bit IO_support flag to 1
setting using_dma to 1 (on)
setting keep_settings to 1 (on)
IO_support = 1 (32-bit)
using_dma = 1 (on)
keepsettings = 1 (on)
root@jac:/#

Is there an advantage to having 'using_dma = 3' as opposed to 'using_dma=1' ?

Also, what about setting the read ahead on ? How did you do that ?

And last, does the 'keepsettings=1' function mean that you only have to set things one time ?
Currently, I have the entry 'hdparm -d 1 -c 1 -k 1 /dev/hda' in my 'rc.local' file so that it sets it every boot.

UnTamed · 12-20-2003, 12:13 PM

Chipset i845GE

Code:

*-cpu
          product: Intel(R) Pentium(R) 4 CPU 2.40GHz
          vendor: Intel Corp.
          version: 15.2.7
          slot: Socket 478
          size: 2736MHz
          clock: 152MHz
*-disk:0
          description: ATA Disk
          product: Maxtor 6Y060L0
          bus info: ide@0:master
          logical name: /dev/hda
          version: YAR41VW0
          serial: Y2RACPJE
          size: 57GB
          capacity: 57GB
          capabilities: ata dma lba iordy smart security pm apm
          configuration: apm=off mode=udma5 smart=on
*-disk:1
          description: ATA Disk
          product: MAXTOR 6L040J2
          bus info: ide@0:slave
          logical name: /dev/hdb
          version: AR1.0400
          serial: 662133433204
          size: 37GB
          capacity: 37GB
          capabilities: ata dma lba iordy smart security pm
          configuration: mode=udma5 smart=on

Code:

>hdparm -Tt /dev/hda /dev/hdb

/dev/hda:
 Timing buffer-cache reads:   2060 MB in  2.00 seconds = 1028.61 MB/sec
 Timing buffered disk reads:  144 MB in  3.04 seconds =  47.34 MB/sec

/dev/hdb:
 Timing buffer-cache reads:   1964 MB in  2.00 seconds = 981.66 MB/sec
 Timing buffered disk reads:  120 MB in  3.00 seconds =  39.98 MB/sec

Never had to mess with hdparm to alter the original settings except when I misconfigured my IDE_BLK_DEV in the kernel.
Same for my older P3/Via133a-ATA66, ...which scores in the wherabouts of Dalek's numbers ...iirc, not booted atm.

lostboy · 12-20-2003, 12:29 PM

Very interesting UnTamed.
I'm going to test mine without setting the flags and see if there is a difference.

You have very impressive scores. I guess that your disks are just a little bit better quality.

How did you get that output that you posted ?

Electro · 12-20-2003, 02:07 PM

hdparm does not benchmark the filesystem. It only benchmarks your hardware. Use bonnie or something similar to find out the real speed for each filesystem.