Test Methodology
NAS Comparisons |
|||||||
---|---|---|---|---|---|---|---|
CPU | Memory | Disk Drives | RAID Mode | Firmware | Date Reviewed | Reviewed Price (Diskless) | |
Qnap TBS-453A-8G NASbook | 1.6GHz Intel Celeron N3150 (quad-core) | 8GB DDR3 | 2x 256GB Transcend MTS800 M.2 SSD (optional) | RAID 1 | 4.2.1 build 20160601 | July 2016 | £500 |
Synology DS216j | 1.0GHz Marvell Armada 385 88F6820 (dual-core) | 512MB DDR3 | 2x 4TB WD Red WD40EFRX (optional) | RAID 1 | 5.2-5644 | March 2016 | £140 |
HEXUS Test Client |
||
---|---|---|
Hardware Components | Product Page | |
Processor | Intel Core i7-6700K (overclocked to 4.4GHz) | intel.com |
CPU Cooler | Noctua NH-D15S | noctua.at |
Motherboard | Asus Z170 Pro | asus.com |
Memory | Crucial Ballistix Sport LT 32GB (2x16GB) DDR4-2400 | crucial.com |
Power Supply | be quiet! Dark Power Pro 11 850W | bequiet.com |
Primary Storage | 256GB Samsung 950 Pro | samsung.com |
Secondary Storage | 512GB SK hynix Canvas SC300 | skhynix.com |
Chassis | Fractal Design Define R5 Windowed | fractal-design.com |
Monitor | Philips Brilliance 4K Ultra HD LED (288P6LJEB/00) | philips.co.uk |
Operating system | Windows 10 (64-bit) | microsoft.com |
HEXUS NAS Benchmark Suite |
||
---|---|---|
Benchmark | Test | Description |
CrystalDiskMark | Sequential Read Q32 | A 1GB sequential read using a 128KB block size at a queue depth of 32. |
Sequential Write Q32 | A 1GB sequential write using a 128KB block size at a queue depth of 32. | |
4K Read Q32 | A random spread of 4KB reads at a queue depth of 32. | |
4K Write Q32 | A random spread of 4KB writes at a queue depth of 32. | |
Intel NASPT | Content Creation | 98 files, 39 per cent sequential, 95 per cent writes up to 64KB |
Office Productivity | 607 files, 81 per cent sequential, 1KB reads and writes. | |
Directory Copy From NAS | 2,833 files, 53 per cent sequential, 64KB reads. | |
Photo Album | 169 files, 80 per cent sequential, 100 per cent reads of various sizes. | |
Power Consumption | Load | NAS power draw recorded while running CrystalDiskMark. |
Idle | NAS power draw recorded when idle. | |
Standby | NAS power draw recorded with drives in standby, if applicable. | |
Noise | A PCE-318 meter is used to record noise output while running CrystalDiskmark. |
Notes
Having made the transition to a new test client, our NAS benchmark results have started afresh with the Qnap TS-453A-8G being compared directly to a mainstream alternative from Synology, the DS216j. The two servers are tested using a single Gigabit Ethernet link with jumbo frames disabled, and are connected to our test client and network via a TP-Link TL-SG1024 switch.
Server and storage benchmarks are fraught with numerous variables, ranging from disks used, file system, network equipment and of course the number of connected clients. Keeping things relatively simple, our tests are carried out using 4TB WD Red (WD40EFRX) hard disks in RAID 1 or RAID 5 configurations depending on number of bays. Alternative disks are only used when a pre-populated NAS is being reviewed.
To provide a look at potential performance, our first benchmark is the freely available and widely used CrystalDiskMark. Our second benchmark - NAS Performance Toolkit, developed by Intel - is a file-system exerciser designed to provide performance comparisons between network storage devices. Intel's utility focuses on user-level performance using real-world workload traces gathered from common applications. NASPT reproduces the file-system traffic observed in various traces onto a chosen storage device and records performance in MB/s.
Last but not least, we measure NAS power draw during three states; under load, idle and standby mode if applicable, while a noise reading is taken using a PCE-318 meter.