The test
We used the same test protocol as in our previous roundup. Here is the material we used:
– five Seagate 500 GB (ST3500320AS) hard drives;
– a Netgear GS-108 gigabit switch which manages Ethernet packets of up to 9 KB (jumbo frames);
– an Enermax Jazz hard drive casing whose particularity is to offer a double USB 2.0/eSATA interface;
– a sound probe, thermometer and ammeter to measure physical parameters.
Tests essentially consisted of timing the copying of the following groups of files in CIFS/SMB and FTP :
– large sized files (two totaling 1.83 GB), two Divx movies;
– medium sized files (63 totaling 1 GB), a mix of MP3 and Flac files
– small sized files (14,746 totaling 659 MB, all inferior to 1 MB), a website !
Finally, the transfer of files was conducted using a machine equipped with the following:
– Core 2 Duo E4500 ;
– 2 GB of DDR2 667 MHz memory ;
– Asus P5K-SE motherboard ;
– two Western Digital WD5000AAKS-0 hard drives in Raid 0 as the source and destination of files;
- a Gigabit D-Link DGE-530T card (PCI, Marvell Yukon circuit).
Performances in Raid 5
Now we come to performances in Raid 5, a mode in which the DS508 is mainly intended to function. We tested storage with 5 drives and then with 4 in order to be able to obtain measurements that are comparable with the competition.
Synology announced more than 30 MB/s in CIFS/SMB writing and we measured… 31 MB/s in sustained speed with a 4 drive storage! Curiously, the same test on the 5 drive configuration resulted in a little more than 27 MB/s. Adding a drive therefore has an impact on the calculation of parity data, which wasn’t the case for Thecus. For this reason, while the DS508 manages to barely surpass the N5200BR Pro in the above mentioned case, the latter is nevertheless more consistent with a little bit less than 31 MB/s in sustained speed in all situations.
In addition, the N5200BR Pro holds a solid lead in the copying of small files with a speed of 5.5 MB/s. This is almost 5 times faster than the DS508 with 4 drives (1.2 MB/s)! Of course, the writing of small files is always problematic and thanks to its Celeron, Thecus’ NAS always do better than the competition. Nevertheless, Synology also does very poorly in this particular test as the CS407 already had results which were largely inferior to those of its rivals. Even the Qnap’s TS-409 Pro has 33% higher performances than the DS508.
In FTP, traditionally faster than CIFS/SMB on large and medium sized files, sustained speeds were coherent with previous results. We gained slightly in speed (+5% à 33.1 MB/s) and we are still faster with a 4 drive storage. However this time, the Thecus N5200BR Pro benefits more from the change to FTP and it surpasses the DS508 in all cases with a minimum of 36.3 Mo/s.
As for small sized files, the DS508 finishes first! This is somewhat incomprehensible because normally the transfer of small files is slower in FTP. However, we already noticed this curious behavior in the CS407 test and it therefore seems to be normal for Synology. On the other hand, compared to the CS407, the DS508 does not improve (with 5 drives) or only modestly (+13% with four drives 4). In the end, this left us a bit disappointed.
And for reading? Here Synology announces 49 MB/s and we measured 52 MB/s on a 4 drive system! With 5, we were back at a sustained speed of 49 MB/s. This is obviously an excellent result and it places the DS508 very far ahead of its rivals. Synology’s new champion proves to be at least 50% faster than the N5200BR Pro!
More good news is that the reading of small files is now a lot less problematic than writing them. The DS508 is thus on the heels of the N5200BR Pro, which despite everything is still the champ (though only a 20% lead). The gap is still somewhat large, but... it is already better than the 400% lead that we saw in writing!
This improvement is something we already saw with the CS407 which attained 3.2 MB/s versus 0.6 MB/s in writing. However, here there is now some real progress with the DS508 as it doubles the speed (6.8 MB/s for a 5 drives system) or even better (7 MB/s with 4 drives).
In FTP, Synology literally explodes out of the starting gates with highs in reading of 72.5 MB/s! For this reason, it proudly takes the lead while Thecus and Qnap are pale in comparison with roughly 45 MB/s (which already were excellent results).
On the other hand, there was no improvement in FTP concerning the reading of small files. In fact, the negociation of each new file transfer seems to take an enormous amount of time in FTP. This is why results already differ greatly when processing 2 large or 63 medium sized files. So with thousands of small ones, performances collapse! The problem is the same in writing but Synology’s performances are so bad in CIFS/SMB that even FTP is faster.
For this reason, it’s a bit of a headache to find the optimal solution for the transfer of files. In fact, it’s better to use FTP in almost all situations, except when it involves the reading of small files where an enormous amount of time is lost compared to a simple Windows transfer. Either way, it’s not very practical to ask ourselves before each operation what kind of files we are sending. Nevertheless, we can establish a few basic rules for the most common types:
- Music and movies: FTP ;
- Photos, various documents (Word, Excel, HTML, etc.) : FTP for writing, SMB for reading.
We can also consider the volume of data rather than the type of files. With dozens of gigabytes, obviously FTP is the better choice. Under one GB, SMB gives us a very good compromise whatever the situation.
Drive failure
What happens when a drive breaks down? To find out, we unplugged a drive during file transfer and then verified the integrity of the file once the transfer was finished. There were no problems to report.
Following a few remarks from some of our readers, we also made this test harder. Instead of unplugging a single drive, we turned off the power during transfer, unplugged the drive and then turned the machine back on. It’s true that drive breakdowns frequently involve power cuts (and therefore the interest of using an uninterruptible power supply). Here again we didn’t have any problems. While data during writing was obviously lost, the integrity of the storage itself was not in danger.
