QNAP TS-559 Pro Turbo NAS
Jul 22nd, 2010 | By AnthonyNAS storage units are heavily dependent on system software, operating systems, network settings and so on. With the built in recorded trace files, NASPT can simulate specified hard drive activity down to the distribution of random and sequential reads/ writes, the spread of data over individual platters, or locality, and a number of access characteristics to ensure consistency between all tests. For our purposes, we will be looking at four tests: HD Video, Office Productivity, Copy To NAS and Copy From NAS.
For clarity, the above image visually represents how a system of disks is accessed, where the vertical axis represents location on the system and the horizontal access represents the total duration elapsed of the test. Our look into IOzone and Iometer have given us an excellent idea of what kind of numbers to expect when it comes to pure or sequential access however real world hardly ever can be judged on theoretical data. The orange and red lines should already be very familiar and are from a recorded trace of moving a single large file and would represent what could almost be near ideal sequential access. The grey line on the other hand is more representative of sequential access in the real world. Were we to take a linear regression of its access pattern, surely it would be similar to either Copy From or Copy To access patterns. Lastly, the mess of the white access pattern is what a real world random access would look like. It is neither entirely random nor entirely sequential, but exhibits traits of each. In Intel’s NASPT, the Office Productivity benchmark models this and is composed of a multiple small but varying sized files stored in different locations and access is comprised of both reading and writing.
Confirming our previous finds, sequential access hovers around the 100 MB/s mark.
Besides the peaks at the beginning of the JBOD and RAID 0 tests due to caching, NASPT also measures write throughput at around and just over 100 MB/s.
But these figures we already know. The last two benchmarks, the HD Video playback and Office Productivity are the numbers we are after.
In our real world sequential read scenario. This test consists of reading a single file approximately 4800 MB in size and for the most part consists of sequential reading. Both the RAID 0 and RAID 5 configurations managed read speeds as we’ve come to expect, however reported transfer rates of around 40 MB/s.
Our last test for NASPT is the office productivity test. Multiple small files, random access and the combination of both read and write is typically the death of most NAS devices (in terms of performance, of course) which pass through our hands. It is really the best true measure of daily usage as a localized storage and I’m pleased to say, the TS-559 Pro managed this better than most averaging just above 40 MB/s.
Before we head over to the last section of testing we are going to finish up our look at NASPT with a simple graph showing average observed throughput during the duration of each test. For the most part the numbers reflect our prior findings, except it should be noted that the calculations of the averages here take in account the systems’ overall cache effect, hence the above possible Gigabit throughput reported in the Copy To NAS section. Negating the data points caused by caching, the TS-559 Pro achieved a read throughput in the mid 80 MB/s range, a write throughput of around 100 MB/s and with random data of small files, just over 40 MB/s.
