|
Synology Cube Station CS407 NAS
|
Author:
Editor:
Sponsor:
Published:
|
Richard Poelling
Max
Synology
Jul. 19, 2007
|
|
|
|
|
|
Testing
Testing was performed through a SMC 8508T Gigabit switch. The test computer uses an A8N-E motherboard (using its onboard gigabit Ethernet) an AMD Athlon 4000+ CPU, 1 GB of Dual channel RAM, and a Hitachi 500GB SATAII drive. Windows XP SP2 had been freshly installed prior to the tests and updated, and the NAS was running with firmware version 2.0.3–0462.
Iometer
For the Iometer tests, the following was performed: 100% sequential distribution, 256 KB transfer request size, either 100% write or 100% read. A maximum disk size of 2048000 Sectors was used. The file that Iometer uses to test is 1,048,576,000 bytes in size. Each test was run for 2 minutes and run 3 times each. A second set of tests was also performed using a 512 KB transfer size.

The RAID 5 jumbo test came back very inconsistent. The spread on the numbers was wide and varied. Even after 4 reps, I was still not able to get the system to calm down.

When I had the Cube Station set on a Jumbo frame, the system seemed to work well, but testing got buggy. Setting the Jumbo Frame size to 9000 yielded a blank test run. Only after setting to 8000 MTU on the Cube Station was I able to get good result. I am not particularly ready to blame the Cube Station outright. Setting a frame size to 1500 on everything was completely stable and showed no problems whatsoever.
The 512K test worked flawlessly, unlike the 256K. Results were exactly what I would expect.
DiskBench
DiskBench version 2.5.0.3 is a small utility which can test such things as file creation and copying.
File Creation
A batch file creation was run from 48 MB to 284MB. I started with 24 blocks of 2MB. Gradually this was increased by 2MB over 60 files. The time it takes to create the file was used to compute the MB/sec speed.

Oddly, the single disk configuration had a better performance than an array. It is doubtful you would want to run in this configuration, because it provides no data redundancy at all. The RAID 5 performance was just fine. For most file sizes, the average difference is about 1 MB/sec or less. This is a performance hit I would be more than willing to take for the increased redundancy that the RAID 5 provides.
Dual File Creation
Two identical files were created at the same time against a mapped drive. They were each 503316480 bytes in size.

For this particular test the addition of the jumbo frame made a considerable difference in the results. In this test, more than 100 seconds were shaved off with the jumbo frame.
File Copy and Read
The file for the Read and Copy test was the ISO file for Windows 2003 Enterprise server. This file is 569366528 bytes in size. I measured both the time it takes to copy the file as well as the time it takes to read the file on the remote system.


The differences between read speeds was minimal; essentially the margin of error for the test. I can’t really say that jumbo frames are a much better enhancement for these particular tests.
|