Introduction
Updated 5/28/2009: Corrected incorrect TS-509 Pro eSATA info.
| At a Glance | |
|---|---|
| Product | QNAP Turbo NAS (TS-809 Pro) |
| Summary | Eight-bay server-like Intel Core 2 Duo-based BYOD NAS with support for single drives, JBOD and RAID 0, 1, 5, 5+ hot spare and 6. |
| Pros | • Online volume expansion and RAID migration • iSCSI target • Local console and remote SSH root access • Built-in LAMP server, DLNA multimedia server, FTP, iTunes • Attached and network backup |
| Cons | • Very expensive • No jumbo frames • No eSATA ports |
This review of QNAP’s TS-809 Pro is the last of QNAP’s current NAS crop that I’ll probably be reviewing for awhile. It’s the largest and most expensive of QNAP’s desktop NASes and can also be had in rackmount form as the TS-809U-RP.
The form factor is basically two TS-439’s stacked vertically, minus a second LCD panel. This makes for a relatively small footprint for a lot of storage.
Figure 1: Front and Rear Panels
Internal Details
Figure 2 shows a view of the 809 from the top with the cover removed. The power supply is easily accessible and the two large case fans can be replaced without dropping the back panel. (Getting at their connectors, however, will require small fingers and some dexterity.) The main board connects to the drive cage backplane via the red SATA cables.
Figure 2: Top inside view
Figure 3 is the left side view, which is dominated by the bottom of the main board. About all you can see here is the backside stiffener for the Core 2 Duo’s heatsink.
Figure 3: Left inside view
Getting the main board out was a bit tricky and after a glance at the components that I could see (Figure 4), I decided to leave the board alone. The design is more similar to the TS-509 Pro than it is the Atom-based TS-439 and 639 Pros, with the main differences being the Core 2 Duo CPU and dual Marvell 88SX7042 PCI-e 4-port SATA-II controllers vs. the 509 Pro’s 1.6 GHz Intel Celeron M 420 and Marvell 88SE6145 SATA controllers. The 809 also has 2 GB of DDR II 667 RAM occupying one slot of two stacked SODIMM sockets, while the 509 comes with 1 GB.
Figure 4: Partial board view
Both systems have two gigabit Ethernet ports provided by Broadcom BCM5787 NetLink Gigabit Ethernet Controllers. This means that the 809 shares the 509’s lack of jumbo frame support. I don’t think this is as much of a drawback as I did when I reviewed the 509 Pro. My experience (and others’ from the SNB Forums) is that the 509 is plenty fast enough without jumbo frames.
Updated 5/28/2009: Corrected incorrect TS-509 Pro eSATA info.
The more serious omission is the 809’s (and 509’s) lack of eSATA ports. This means that throughput to external USB drives for either backup or storage expansion is limited to 23 MB/s or so. In contrast, even the Atom-powered TS-439 and 639 Pros were able to achieve backup speed in the mid-to-high 60 MB/s to an EXT3-formatted eSATA drive.
With eight Samsung HE103UJ 1 TB drives provided by QNAP with the review unit spun up, the 809’s power consumption measured 102 W. With the drives spun down for power saving, wattage dropped to 47 W.
The 809 was almost shockingly quiet in operation when you consider the number of spindles whirring away. But you will, of course, know that it’s running if you have a quiet room.
Performance
The 809 was tested with our standard NAS test processes. I upgraded to the latest 2.1.6 Build 0428T and used the Samsung HE103UJ 1 TB drives installed by QNAP. Tests were run with four drives configured in RAID 0 and 5 with a 1000 Mbps LAN connection.
Backup
I first checked backup performance with the results summarized in Table 1 along with the TS-509 Pro’s for comparison. Best performance is surprisingly with the USB drive NTFS-formatted! You can check the 809’s backup rankings for FAT and EXT3 formats, where it ranked fourth and sixth.
| Product | QNAP TS-809 Pro Backup Throughput (MBytes/s) |
QNAP TS-509 Pro Backup Throughput (MBytes/s) |
|---|---|---|
| USB – FAT32 | 19.8 | 19.2 |
| USB – EXT3 | 20.3 | 24.1 |
| USB – NTFS | 22.7 | 21.3 |
Table 1: Attached backup throughput test summary
I also tested Remote Replication for NAS-to-NAS backup to a TS-439 Pro. Table 2 summarizes the results for full backups of the test folder without and with the encryption option (SSH connection) enabled, which really slows things down.
| Product | QNAP TS-809 Pro Backup Throughput to QNAP TS-439 Pro target (MBytes/s) |
QNAP TS-439 Pro Backup Throughput to QNAP TS-809 Pro target (MBytes/s) |
|---|---|---|
| NAS-NAS | 31.89 | 33.07 |
| NAS-NAS (SSH encrypted) |
10.24 | 11.16 |
Table 2: Networked backup throughput test summary
Benchmarks
Figure 5 presents a summary of the benchmark tests run for the 809 with RAID 0 and 5 write and read plotted. Performance boost at the smaller file sizes isn’t extreme for either RAID 0 or 5 writes and read performance is pretty flat across all file sizes tested. As with the TS-439 Pro, you don’t give up a lot of speed by using RAID 5 vs. RAID 0.
Figure 5: Performance benchmark summary
RAID 0 and 5 Performance with a 1000 Mbps LAN connection averaged over the 32 MB to 4 GB file sizes and with cached results above 125 MB/s removed from the average is summarized in Table 3. The 809’s lower ranking in the RAID 0 write test is due to the effects of removing cached performance from the calculations.
| Product | QNAP TS-809 Pro Average Throughput (MBytes/s) |
Chart Rank |
|---|---|---|
| RAID 0 Write | 74.3 | 4 |
| RAID 0 Read | 84.5 | 1 |
| RAID 5 Write | 65.8 | 1 |
| RAID 5 Read | 80 | 1 |
Table 3: Average throughput summary
Performance – Competitive
Since the 809 stands alone as an eight bay NAS (at least among the products that I have tested), I’ll have to compare it against six and one seven bay competitors: NETGEAR’s ReadyNAS Pro, QNAP’s TS-639 Pro and Thecus’ N7700.
For RAID 5 write (Figure 6), the products clearly cluster in two groups, the higher-performing TS-809 Pro and Thcus N7700 and the (relatively) lower performing ReadyNAS Pro and TS-639 Pro. At the largest file sizes, there is around 20 MB/s separating the two groups!
Figure 6: Competitive RAID 5 write comparison – 1000 Mbps LAN
For RAID 5 read (Figure 7), the 809 actually comes out on top across all "large" file sizes. Note the big drop in the N7700’s throughput above 512 MB.
Figure 7: Competitive RAID 5 read comparison – 1000 Mbps LAN
Performance – File Copy
The Vista SP1 1000 Mbps RAID 5 file copy write chart (Figure 8) shows the 809 clearly out-performing the other products at 90.4 MB/s.
Figure 8: 1000 Mbps LAN Vista SP1 File Copy Write
Not so much for RAID 5 filecopy read (Figure 9), however, where the 809 is tied for 3rd / 4th place with the TS-439 Pro at around 86 MB/s.
Figure 9: 1000 Mbps LAN Vista SP1 File Copy Read
Use the NAS Charts to further explore performance.
iSCSI
The HD Tune Pro tests I used in previous iSCSI tests weren’t doing much for me and I had to decide on something to use in order to create iSCSI Write and Read performance NAS charts. So I settled on using the same file copy test used in the NAS-connected via CIFS/SMB.
Figure 10 shows the Thecus handily beating all the QNAP NASes and the NETGEAR ReadyNAS NVX. The ReadyNAS Pro isn’t shown because it only recently added iSCSI support and I have long since sent the Pro back to NETGEAR.
Figure 10: iSCSI Write
The 809 did much better with iSCSI read, topping the chart at almost 94 MB/s and surpassing the N7700.
Figure 11: iSCSI read
Link Aggregation
NETGEAR has been kind enough to loan me a GS108T ProSafe 8 Port Gigabit Smart Switch so that I can test link aggregation in high-performance dual-ported NASes that support it. Since the 809 fits that description, I ran a quick test.
I first set the 809’s Ethernet ports to "load balance" mode, which is supposed to support 802.3ad link aggregation. I then added the two GS108T ports connected to the 809’s ports to a Link Aggregation group, then enabled LAG in the GS108T.
I then ran Vista Filecopy write and read tests using the NAS Testbed and another Core 2 Duo system running Vista SP1, but only a single drive (not a dual-drive RAID 0 array as on the NAS test bed machine.)
I first ran the test on each system individually, then with both systems running simultaneously. If aggregation is working properly and if the NAS under test is being limited by the capacity of a single Gigabit Ethernet connection, enabling link aggregation should allow higher net throughput. The test results are summarized in Table 4.
| Product | NAS Test Bed Throughput (MB/s) |
System 2 Throughput (MB/s) |
Total Aggregate Throughput (MB/s) |
|---|---|---|---|
| File Copy Write | 92.4 | 57.4 | 149.8 |
| File Copy Read | 86.7 | 55.8 | 142.5 |
| Simultaneous File Copy Write |
56.3 | 41 | 97.3 |
| Simultaneous File Copy Read |
30 | 32 | 62 |
Table 4: Link Aggregation Test
The results show no improvement from using link aggregation. The total write capacity of the two computers is just shy of 150 MB/s. But when both systems wrote to the 809 with aggregation enabled, throughput dropped for both systems and totaled only 97 MB/s. This is only 5 MB/s higher than the highest individual system write, which is within the margin of error of the test method.
Read was a similar story, with only 67 MB/s of total throughput out of a possible 97 MB/s achieved with both systems running simultaneously. This is again, only a slight (6 MB/s) gain over the highest individual system read speed.
I’ll await feedback from QNAP and rerun the test if they have suggestions on my methodology. But, for now, it doesn’t look like link aggregation will buy you very much performance improvement with the 809.
Closing Thoughts
At $1700, the TS-809 Pro is not a product that most consumers will look twice at. But the performance results are perhaps useful for those DIYers who are basing their homebrew NASes on Intel’s Core 2 Duo and want to know what can be achieved.
I suspect most consumer NAS buyers set on a QNAP product will be better served (HA!) by opting for one of the Atom-based models. Not only will you get a lower price, but also eSATA ports for faster backup and jumbo frames to play with for possible performance improvement.