| QNAP TS-259 Pro Turbo NAS Server |  |
| Reviews - Featured Reviews: Network | |
| Written by Bruce Normann | |
| Sunday, 02 May 2010 | |
|
Page 10 of 13
1GB Single-Disk Test ResultsThe bottom line for any storage device is the combination of capacity and transfer speed. For a network attached storage server, the differences are all about the infrastructure that is placed around the basic HDD array. Since capacity is something that's easy to define and measure, the real question for any NAS product is how fast will it Read and Write data. For this reason, Benchmark Reviews measures NAS performance as the bandwidth achieved during a file transfer test. The first tests we perform utilize a single 1GB (1000 megabytes / 1,000,000,000 bytes) file in a transfer to and from the NAS. QNAP pulls down the top five spots for 1GB Read testing, with their TS-xxx series of products. The TS-809 Pro and TS-509 Pro are built from similar hardware, and the TS-459 and TS-259 are also kin under the skin. Even without the extra Jumbo packet size, the QNAP x09 products are built for very high transfer rates. At 62.5 MB/s read speed (using EXT4) the TS-809 Pro beats everything except the QNAP TS-459 Pro with 9k Jumbo Frame enabled. Our two-bay TS-259 Pro comes in slightly below the TS-509 when using a standard MTU of 1500, and then matches it with a transfer rate of 59.0 MB/s using 4k Jumbo Frames. The single-drive QNAP TS-119 and Synology DS408 come in close with Jumbo Frames enabled, reaching 55.6 MB/s each. The Thecus N7700, Synology DS209, and QNAP TS-419P come in right behind the leaders, with a matching best effort of 50.0 MB/s. Although the QNAP TS-419P shares the same hardware as the TS-119, for some reason it didn't post the same performance numbers with Jumbo Frame support.
Moving on to the 1 GB write bandwidth test, our results suggest that while it may sometimes it will be faster to read files from a NAS server then it is to write files onto it, the opposite is true more often than not. The Network Attached Storage servers each have their own strengths, with some being more robust in write performance while others have more impressive read performance. At the top of our charts the QNAP TS-809 Pro offers outstanding performance again (with EXT4), leading the crowd even more than it did in the 1GB read tests, with 76.9 MB/s of write speed. Formatted with EXT3, the TS-809 Pro offered 71.4 MB/s, followed by the QNAP TS-509 Pro which maintained 62.5 MB/s without the aid of Jumbo Frame support. Next in line, the Thecus N7700 does well in write tests and produced a best result of 62.5 MB/s, finishing out the list of top-performers. The QNAP TS-259 leads the next grouping, with write speeds of 43.5 MB/s using a standard MTU of 1500, and 44.8 MB/s with a Jumbo Frame size of 4074. It narrowly beat the QNAP TS-459 Pro NAS, which produced 41.7 MB/s at 1500 MTU, and remained relatively unchanged with 41.8 MB/s once 9000 MTU Jumbo Frame was enabled at each end of the network connection. The Synology Disk Station DS209 narrowly beat the larger DS408, and scored a swift 43.5 MB/s write performance with Jumbo Frame enabled. A single-drive QNAP TS-119 held a steady 33.3 MB/s speed, scoring the same with and without Jumbo Frame enabled, and was tailed by the Thecus N3300 at 32.3 MB/s.
The lower-end group consists of all the remaining NAS products, with the next closest competitor being the QNAP TS-419P which scored 25.6 MB/s. The remainder of the bunch trailed far behind. Next up is 10 GB (1000 metric megabytes / 10,000,000,000 bytes) testing. Using the single-disk configuration in each NAS, and a single Gigabit connection, network throughput will be put to the test. 10GB Single-Disk Test ResultsExamining 10GB basic file transfer speeds, the QNAP TS-459 Pro delivers remarkable read performance, at 65.4 MB/s with Jumbo Frame enabled. QNAP's newer models, the TS-459 Pro and TS-259 Pro lead the field on Read results, based on their Jumbo Frame capability. Both of them, plus the TS-809 Pro, and TS-509 Pro are in a dead heat utilizing the standard MTU value; all four of them are within spitting distance of 56 MB/s. QNAP's single-drive TS-119 offered the same performance once Jumbo Frame was enabled, but is reduced to 47.6 MB/s at 1500 MTU. Once again, in Read performance, QNAP takes the top five spots on the chart. The Synology DS209 and DS408 maintain a close relationship with peak performance around 47 MB/s in normal mode, and improve to 49.3 and 48.3 MB/s read performance with Jumbo Frame enabled at both ends. At 50.3 MB/s with Jumbo Frame support, the QNAP TS-419P offers very good large file transfer speeds, but without JF support the speed slows to 39.8 MB/s. The Thecus N7700 was able to sustain 44.5 MB/s using Jumbo Frame, and 41.5 MB/s without it, which was very similar to the N3200 Pro.
In our 10GB write performance tests, the charts get shuffled a bit and the results also take on a more linear profile, instead of a series of groupings. The Thecus N7700 makes up for above-average read performance by achieving the top result for large-file writes, with 56.2/58.8 MB/s in normal and Jumbo Frame modes; for once QNAP was not the performance leader. Next in line comes the QNAP TS-809 Pro, with 54.4 MB/s using EXT3. Then the TS-509 Pro, which scores 47.0 MB/s. The QNAP TS-259 Pro slots in next with 42.9 MB/s and 44.6 MB/s with 4k Jumbo Frame enabled, rising above the Synology DS209 and the QNAP TS-459 Pro just like it did in the 1GB test. The TS-459 drops to 39.2 MB/s at 1500 MTU and 40 MB/s with Jumbo Frame enabled, just losing one position to the Synology DS209 in this test. The remainder of the products trail off in a very linear fashion, until you get to the Patriot Corza, which really struggled with both of the 10GB tests; then again this network appliance costs less than $100.
This is the end of our test results for this review, since the QNAP TS-259 does not support the RAID5 configuration we normally use to test the large format NAS products. A couple of things struck me about the results. The performance you can expect is based largely on the infrastructure that surrounds the hard drives in all these products. The processing power, amount of system DRAM, network interfaces, and configuration options all make a significant difference. This is all the more surprising since the single-disk configuration is the least demanding in terms of the resources driving it. Any disk configuration of RAID-5 or higher introduces a much greater load on the supporting subsystems of a NAS device. If you plan to deploy one of these devices in a multi-user setting, you also have to account for the simple fact that 2-3 people are probably going to hit the server at the same time, usually just before they have to leave to catch their train home at the end of the day. So, it pays to think of these systems as servers first, and disk storage arrays second. In our next section, Benchmark Reviews looks at the surprising differences between the Windows XP and Windowa 7 environments. Be sure to take a look, especially if you would like to find out which platform performs 50% better... NAS Comparison Products
|
|
Comments
I think the way to test this feature is to have two or more transfers occurring at the same time. With one transfer already under way, another could be started and timed, and the speed of the second transfer should be relatively unaffected by the continued activity of the first one. Your thoughts, suggestions?
If you can read french, I posted a lot about it on #lafibre.info/iperf/gs108t-nc360t-n5550-load-balancing-33mbs/new/#new but the result is this one :
Conclusion is 802.3ad is ONLY failover. ABSOLUTELY NOT load balancing.
If you want to do what I dreamed of, choose on each side the Balance-SLB (or Balance-ALB) + round robin transmit load balancing method.
Then you will have a smooth repartition of the packets on each port, but you will notice a strong down bandwidth due to the fact that "Packet order is NOT guaranteed"
Load balancing for a one file transfer on many cables is just a dream... right now...