|QNAP TS-870U-RP NAS Network Storage Server|
|Reviews - Featured Reviews: Network|
|Written by Bruce Normann|
|Tuesday, 04 June 2013|
Page 14 of 17
Intel NASPT Test Results
NASPT brings an important perspective to our test protocol, as it is designed to measure the performance of a NAS system, as viewed from the end user's perspective. Benchmarks like ATTO use Direct I/O Access to accurately measure disk performance with minimal influence from the OS and the host platform. This provides important, objective data that can be used to measure raw, physical performance. While it's critical to measure the base performance, it's also important to quantify what you can expect using real-world applications, and that's exactly what NASPT does. In keeping with the real-world scenario, I only run these tests on the RAID 5 configurations, since that is what most users with a mid-size NAS are going to use. It just doesn't make sense to run realistic test scenarios on unrealistic hardware configurations.
One of the disadvantages of NASPT is that it is influenced by the amount of memory installed on the client, and it was designed for systems that had 2-4 GB of RAM. Consequently, two of the tests give unrealistic results with modern systems, because they are measuring the speed of the buffer on the client, instead of the actual NAS performance. For that reason, we will completely ignore the results for "HD Video Record" and "File Copy to NAS". You can a batch run of 5 cycles through the tests, but my results turned out to be a bit slower than the individual runs. There seemed to be some wrinkles in the batch testing that don't show up on individual test runs, which is a bit of a pain, to be honest. The numbers in the chart below are an average of five separate runs, which I believe are more accurate than results from a consolidated batch run.
With a single, 10GbE interface in place, the results look quite a bit better than any set of data I have from NAS models with the slower GbE interface. Results for several of the tests that are more sequential in nature are in the 200-300 MB/s range. Some of the tests have very low transfer rates, and that's due to the nature of the test. The Content Creation test for example, simulates a user creating a web page, accessing multiple sources for the content. The Directory Copy tests use several hundred directories and several thousand files to test a typical backup and restore scenario. That's one of the most real-world types of test, and it's useful for all of us to have a standard set of test data to use, because my directory of 1,000 random small files is never going to be the same as your directory of 1,000 random small files.
To summarize things, here are consolidated charts of the "Fast" NASPT tests, the "Medium-Speed" tests, and finally the two "Slow" ones. I've regrouped these into four charts this time, in order to make them more legible. As the number of entries grows, the text gets too small to read, at some point. First, the two tests with the highest transfer rates. The TS-870U-RP didn't fare as well on the HD Video Playback trials as I expected it to. I always looked at these two benchmarks as essentially sequential Read tests, so the star Read performance of the TS-870 should really be surfacing here. With 4x Video, the TS-870U-RP hits a high of 274.1 MB/s, with the 4-disk configuration, and only scores 235 MB/s with all eight bays loaded up. The TS-879U-RP, with the faster CPU, runs past it at 457.6 MB/s in the same 4x HD Video benchmark.
The Medium-Speed tests show the TS-879U-RP putting in another very strong second place finish to the TS-879U-RP with the faster CPU. The higher performance with the four-disk array continues here, especially in the File Copy test. My file copy tests certainly didn't show this trait, so I don't know what's causing it. The EonNAS 850X takes third place in this grouping, and all the NAS units with single GbE interfaces are left in the weeds. The mixture of Reads and Writes makes this set of charts a little topsy-turvy.
The "Slow" tests generally are slow because the file transfers are done with data sets that contain a bunch of small files of irregular size. In addition, the Directory Copy tests are accessing the file system index much more heavily than in the other tests. This adds a unique component that could be critically important for some users. The Directory Copy To NAS and Directory Copy From NAS results show once again the effect of the 10GbE interface on performance. The top three contenders stay far ahead of all the other NAS solutions, with one difference; the EonNAS 850X takes top honors for the first time, no doubt due to the influence of its Solaris-based operating system. Its ZFS file system is known for being a monster, and it shows that quality here. This is a good demonstration of why it pays to look closely at your potential use cases when shopping for any H/W or S/W solution. The TS-870U-RP puts in a second place effort, beating out the more powerful TS-879U-RP by enough of a margin the Read test that it compensates for the smaller lead that the TS-879U-RP takes in the Write test.
The Office Productivity and Content Creation are my least favorite tests in the NASPT suite for two reasons. One, I don't see most NAS products being used in this fashion. Maybe I'm behind the times, and desktop virtualization will make this a very important benchmark in the near future. Second, the results never seemed to scale very well with the system performance that I was seeing on other tests. In any case, the TS-870U-RP turned in some very unusual results on this particular test. With four disks in place, in RAID 5, it swept the field with a record result of 72.2 MB/s. With eight disks, it came in a very sorry last place. There seemed to be large gaps with no network activity, as I watched the bytes go by in the host computer's performance monitor, and I suspect this is why the average result is so low. Ignoring this anomaly, the top three performers in the rest of the tests, the two 8-bay QNAPs and the EonNAS 850X, still held on to the top three spots in the rankings.
The Photo Album test is a bunch of small files again, of varying sizes, arranged in a complex directory structure. This is a very common type of data set, and these results were roughly similar to the previous test, except the 8-disk configuration on the TS-870U-RP didn't fall so far behind. The four-disk configuration won the test outright, again.
The NASPT benchmark showed some real world anomalies this time, which I wasn't expecting. Beyond the simple sequential results that are easy to measure and very consistent, is a world of data that is immeasurably complex in its variations. The Intel NASPT suite is one of the few that challenges NAS devices with realistic data sets. The results can be a bit messy to interpret, like they were in this case, but they've also provided the greatest potential insight into NAS performance of any commonly used benchmark. Stay tuned to see if we have similar challenges on some non-traditional NAS tests, in the next section.
NAS Comparison Products