DS212J Performance on RAID-1

Ho, ho, ho!

X-mas is here and Hogfather has left a nifty present before my fireplace. The Synology DS212J is a new sibling in the family and comes to replace a DS211J, the latter shining with its speed-to-price ratio, offering SMB/CIFS read/write at almost 50MB/s for less than $120. Now the new generation is supporting more hardware drives, including another favorite of mine – the WD20EARX, offering a 6G SATA port plus smart caching and data management, which can effectively offer up to 128MB/s data bursts in a very quiet chassis and at an opportune price.

Figure 1: WD20EARX performance, courtesy of http://img204.imageshack.us/img204/4562/wd20earxint.jpg

The hard disks were recognized from the Synology NAS and S.M.A.R.T.-info and -test can be easily retrieved from the Web GUI in the DSM 3.2 version.

Figure 2: HDD Management under DSM 3.2

I configured the disks as RAID-1 that ensures high availability at a moderate performance loss and decided to measure the transfer speed for the SMB and iSCSI protocols. AppleTalk and NFS are not tested here, sorry.

For this purpose I used a widespread nice-and-easy software tool ATTO in version 2.34. The current version 2.47 is not supporting mapped network shares anymore. Also I abstained from using other tools, since – as you will see later – even the ATTO results were very positive, compared to the actual user experience.

So let’s go ahead and dive into the numbers:

Here we are measuring a SMB share, mounted on a Windows 7 as Z:

Figure 3: SMB/CIFS Share [-Z-]

As you can see the performance is miserable for everything fewer than 64KB, after that stabilizing to roughly 65MB/s write and 73MB/s read speed averages. As I could fast check there was no bottleneck; neither on the transfer media (cables, switch), nor on the CPU or RAM utilization of the NAS as depicted below:

Figure 4: Synology resources during the SMB speed test

On first thought I would presume that the performance loss is due to the RAID-1 configuration but on the other hand I find the difference from 120MB/s to 80MB/s too large a victim for a single Linux /dev/mdX driver issue. It will be interesting to compare my outcomes with other independent results…

The second test scenario was a 2GB large virtual disk file on the DS212J mapped as and iSCSI target and mounted on the Windows 7 workstation as drive H:

Figure 5: iSCSI Share [-H-]

Here the results are even more sobering; the “stable” averages starting first with 128KB large data blocks are efficiently showing us only 45MB/s write and 58MB/s read speeds.

Once again here the resources were not stressed, with an average CPU-workload of approx. 60%:

Figure 6: Synology resources during the iSCSI speed test

The reader may draw the conclusions for his own…

The last test iteration was including the “real user experience”.

First we started with copying big chunks of data (~700MB ISO images) from a local SSD with an average read speed of 520MB/s for large files:

Figure 7: OCZ RevoDrive 80GB PCIe x4

…to the mapped SMB/CIFS network share…

Figure 8: Copy large files with Windows Explorer

…and during this observed the CPU utilization…

Figure 9: Synology resources during the Windows Explorer Copy test

Also the small files of 51KB average size were not very efficiently copied.

Figure 10: Copy small files with Windows Explorer

These results are somewhat disappointing compared to the ATTO tests, especially the fact that I never achieved a max speed higher than 55MB/s, still utilizing the NAS CPU to bold 71%.


On the one hand the NAS appliance is very quiet and offers a great amount of features and options. Also for the private users or for home offices it offers acceptable performance characteristics – especially for people transferring from 100 Mb/s network environments. On the other side I cannot conceal some negative experiences on the network share protocols and especially on the iSCSI implementation which can be very useful for backup scenarios on small offices – since the firmware is essentially the same for all models, including such as four-bay NAS appliances. In my opinion there are many performance improvements that can be implemented in future Synology firmware releases.

I will keep my new baby and hope for new firmware releases, healing the side effects that I experiences during this hasty examination.