NAS Configuration

Test Setup for Single-Client Tests

The test system (client PC) for single-client tests consists of the following components:

Methodology

As we're starting fresh with NAS reviews, the benchmarks have changed quite significantly. We're running a set of synthetic benchmarks to measure the theoretical performance to start with and then a set of sequential data copying benchmarks to get some real world performance figures. The synthetic benchmarks consist of CrystalDiskMark, iPerf3, NAS performance tester and Helios LanTest.


CrystalDiskMark for those that aren't familiar with the benchmark, is a popular Japanese drive benchmark that tests sequential and random read and write performance of a drive. We simply map a network drive in Windows and run CrystalDiskMark on the mapped drive.

iPerf3 is a command line network throughput tests that is a simple, yet powerful way of testing network throughput. Although it's possible to perform drive performance tests with iPerf, it's normally only used for testing raw network throughput, which is what we use it for in our NAS reviews. It's also a good way to make sure that there are no network configuration issues which would cause slowdowns of any kind.


The NAS performance tester is an open source tool for testing NAS appliances. It creates a file that is written or read from the NAS, then deleted to prevent any kind of caching happening, as normally this test is looped at least five times and an average is calculated of the five runs. We chose to use a large data set of 8 GB here, which puts some strain on the NAS under testing.


The Helios LanTest is another NAS benchmarking tool, but this time around we're looking at a tool that performs multiple tests. It has five response time tests of increasing difficulty as well as two sequential performance tests. It also has a default setting for 10 Gbps Ethernet, which we use for testing 2.5 Gbps and 5 Gbps Ethernet, as the 1 Gbps test is simply not enough of a challenge. In case of the responsiveness/latency tests, a lower number is always better, and obviously it's the opposite for the sequential read and write tests.

This is then followed by three data copying tests. The first of these three tests is a game library that consists of 10,458 files for a total of 96.2 GB worth of data. The second test is a single 66.9 GB file and the third test is 4,084 photos which in total adds up to 10.6 GB worth of data. These files are copied from an SSD to the NAS and then from the NAS back to an SSD. This gives us a result in minutes and seconds that is then converted to an average MB/s transfer rate. Note that the actual transfer rate can vary a lot during these tests, especially in the first test due to both very large and very small files being copied.

In addition to this, we also tested the NASync DXP4800 Plus with an SSD installed as a read cache. As this had zero effect on the synthetic tests, we only ran the file copy tests here. Each test was performed five times to allow the SSD to cache enough data and the results are from the fifth test run.