DSM-G600, DNS-3xx and NSA-220 Hack Forum

my results:

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index
                    :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          127.36  :       3.27  :       1.07
STRING SORT         :          9.3477  :       4.18  :       0.65
BITFIELD            :      2.9155e+07  :       5.00  :       1.04
FP EMULATION        :          8.9321  :       4.29  :       0.99
FOURIER             :          27.395  :       0.03  :       0.02
ASSIGNMENT          :          1.3099  :       4.98  :       1.29
IDEA                :          365.53  :       5.59  :       1.66
HUFFMAN             :          77.131  :       2.14  :       0.68
NEURAL NET          :        0.041329  :       0.07  :       0.03
LU DECOMPOSITION    :          1.2549  :       0.07  :       0.05
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX       : 4.001
FLOATING-POINT INDEX: 0.062
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU                 :
L2 Cache            :
OS                  : Linux 2.6.6-arm2
C compiler          : gcc version 3.3.5 (Debian 1:3.3.5-13)
libc                : ld-2.3.2.so
MEMORY INDEX        : 1.000
INTEGER INDEX       : 1.000
FLOATING-POINT INDEX: 0.031
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.

Here's an overview of the standardized units: http://physics.nist.gov/cuu/Units/binary.html

MB = 1.000.000 bytes
MiB = 1.024 * 1.024 bytes
Mbit = 1.000.000 bits
...

phoenix wrote:

Not bad for a networked drive, I'd say. An USB-attached drive shouldn't be much faster, as we're at the mechanical limits of the disks, I think.

Wikipedia has a chart http://en.wikipedia.org/wiki/Serial_ATA … ther_buses which lists various hard drive bus technologies as well as the SATA II specification.   15-20MB/sec is well below "the mechanical limits of the disks".

There is a web site smallNetBuilder http://www.smallnetbuilder.com/content/ … 71/75/1/7/ that has a review of the DNS-323 on page 8 of the review they list performance specification, and a description of how snallNetBuilder tested the DNS-323 system.

Using the smallNetBuilder test format, I get ~15MegaBytes/sec and ~20MegaBytes/sec read on a gigabit network from a winXP machine using the default samba (DNS specs in my signature).

Last edited by mig (2007-05-12 02:29:48)

configuration:

DNS-323 with Hitachi Deskstar T7K250 250GB (formatted on EXT2) firmware 1.01 (factory)
D-Link DGS-1005D - Switch Gbit
External HDD on usb 2.0

transferring over 100GB of multimedia files about 700MB each files from usb2 drive to DNS-323

i registered a transfer rate of (writing to DNS-323) of 140Mb/s (read from i-band usr)
the flow work from a minimum of 135 to a maximun of 143 Mb/s

Last edited by raoul.grasselli (2007-05-11 11:52:10)

What is everyone using to test their network transfer speeds? I kind want something that will be comparable across the 3 different types of machines

I have 2 vista PCs, 3 XP machines & 1 OSX machine in my house and I am trying to determine their transfer speeds based on where they are located in the house.

I am trying to do this to determine if the current transfer slowdowns are caused by crosstalk in the cabling based on its location.

thx

Let's raise an old topic.

While upgrading my disks, I did a little throughput test to determine optimum setting for my use. I hope others find my results interesting.

I run an all linux network. Sometimes I boot to Xp for gaming. Only performance issues so far had been when watching bluray images over network, so I tested throughput of single 1GB file. I also updated firmware to 1.08beta8. First two questions were if enabling jumbo frames would give a performace boost, and which network filesystem, nfs or samba, is faster. All test were run on ext3 filesystem with disks as two seperate volumes.

nfs w/ jumbo frames    read 18,4 MB/s, write 6,9 MB/s
nfs w/o jumbo frames    read 19,7 MB/s, write 12,1 MB/s
samba w/ jumbo frames    read 12,4 MB/s, write 837 kB/s
samba w/o jumbo frames    read 12,9 MB/s, write 834 kB/s

Poor samba performance on writes were likely caused by a bug in linux CIFS driver. On XP I got about 16 MB/s read and 13 MB/s write with samba. Negative impact of jumbo frames were propably because of buggy desktop hardware or router. Enabling jumbo frames on desktop PC caused file transfers to hang, so jumbo frames were enabled only on DNS323. On a network with all hardware supporting jumbo frames results might have been different. I decided to go with nfs and without jumbo frames enabled.

Next question was, what kind of performace impact does ext3 filesystem have on DNS-323 in comparison to ext2.

Ext2    read 20,0 MB/s, write 15,4 MB/s
Ext3    read 19,7 MB/s, write 12,1 MB/s

Last question was, how much does performace change when disks are in two seperate volumes, JBOD, Raid0 or Raid1 configurations. All test were run on ext2 filesystem.

2 vols    read 20,0 MB/s, write 15,4 MB/s
JBOD    read 19,0 MB/s, write 15,5 MB/s
Raid0    read 20,9 MB/s, write 15,3 MB/s
Raid1    read 19,4 MB/s, write 13,5 MB/s

After these results I decided to go with nfs on ext3 formated drives configured on two seperate volumes.

i saw i syncing raid1 with 80-100 MB/s. But over samba i only get 15 MB/s. I can live with that.

anodyne wrote:

Did some testing for my own interest to find the performance hit of EXT3 vs EXT2. I also decided to compare individual disk vs RAID0 & RAID1. And I did it with two types of workload: one large file (~400MB) and multiple (10000+) small files (~900MB total). This is with a 2.66GHz P4 system (PATA disks) running XP on a gigabit network, standard frames. I'm connecting to the DNS via Samba. Here are the summarised results (values are transfer rates in MB/sec):

1) Single large file
                        Individual   RAID0    RAID1
Ext2 filesystem     12.4       12.7      12.1
Ext3 filesystem     10.4       10.2        9.5

2) Multiple small files
                        Individual   RAID0    RAID1
Ext2 filesystem     2.5          2.5         2.4
Ext3 filesystem     2.3          2.3         2.2

Conclusion: in this environment, EXT3 is 10-20% slower. RAID1 has a slight performance penalty. And there's no benefit to RAID0.

Well done!!!

Well thought through and well written - many people have done their own tests, but the test methodology and the write up were so poor, that the numbers were not really suited to anything other than a "mine is faster than yours' comparision.

I note you say you were connecting to the DNS via SAMBA and interpret this to mean that you were running linux on the P4 system - I vaguely recall a couple of threads that suggest the DNS-323 does not perform as well in this environment as it does in a Windows environment.