Testing (continued):

Real World File Transfers - Large Files:

In this phase of testing we will look at the speeds during real world file transfers. For these tests, the files described were transferred between the Windows desktop PC and one of the storage servers while a stopwatch was used to record the transfer time. The file size and the transfer times were then used to calculate a transfer rate.

Large file transfer rates were the first type to be tested, and we will look at small file transfers on the next page. Disk images, HD video files, and large multimedia projects would fit in to this grouping. The large file transfer test involved moving a single 5.96GB (6103MB) MPG video file around the network to the storage servers of interest. A reboot of the Windows desktop system was executed between each test, and different folders were used for reading and writing.

When reading the large file from the DX4200 we see some very impressive transfer rates - no matter what storage layout was chosen it was right around 110MB/s, which is the fastest of all devices on hand.



On the write side of the large file test we something interesting develop, as the parity layout is much slower than simple or mirror. Simple and mirror layouts offer write transfer rates that nearly match the read transfer rates - which are definitely impressive. Unfortunately, the parity test is the one I ran first and the results made me concerned that something was wrong - so I contacted WD. After an informative conversation I learned that parity results won't look as good as simple or mirror, and that what I was seeing was normal. Considering parity is most like RAID 5, this would be my preferred configuration, but not if all write tests looked like this.