How We Test Surveillence HDDs
Test Hardware | |
---|---|
Processor | Intel Core i7-3960X (Sandy Bridge-E), 32 nm, 3.3 GHz, LGA 2011, 15 MB Shared L3, Turbo Boost Enabled |
Motherboard | Intel DX79SI, X79 Express |
Memory | G.Skill Ripjaws Z-Series (4 x 4 GB) DDR3-1600 @ DDR3-1600, 1.5 V |
System Drive | Intel SSD 320 160 GB SATA 3Gb/s |
Host Bus Adapter | LSI SAS 9300-8e |
Tested Drives | WD Purple 4 TB: WD40PURXSeagate Surveillance HDD 4 TB: ST4000VX000 |
Comparison Drives | WD Red 4 TB: WD40EFRXWD Se 4 TB: WD4000F9YZWD Re 4 TB: WD4000FYYZSeagate Savvio 15K.3 146 GB: ST9146853SSSeagate Enterprise Performance 10K HDD v7 1.2 TB: ST1200MM0017Seagate Constellation.2 1 TB: ST91000640NS |
Graphics | AMD FirePro V4800 1 GB |
Power Supply | OCZ ModXStream Pro 700 W |
System Software and Drivers | |
Operating System | Windows 7 x64 Ultimate |
DirectX | DirectX 11 |
Driver | Graphics: AMD 8.883 |
Benchmark Suite | |
HD Tune 2.55 | Block = 8 MB |
WD Surveillance Benchmark | 16 Cameras x 720p x 30 FPS16 Cameras x 1080p x 20 FPS32 Cameras x 720p x 30 FPS |
Testing a surveillance drive should be different than running a normal consumer or enterprise drive through its paces. Because the workload is so narrowly defined, we chose only tests that would mimic a surveillance environment. As a result, you won't see typical consumer or enterprise tests, such as SYSmark or enterprise server profiles in Iometer.
One test that we did include was provided to us by Western Digital. You know that we wouldn't accept tests or scripts directly from manufacturers without fully vetting them, and we've seen enough from this one that the results appear in line with other tests and our expectations. It should be stated, however, that WD's Surveillance Benchmark directly aligns with the type of transfers that WD is targeting with the Purple. The total workload is calculated by using the following equation:
Total Workload = (# of cameras) x (bit rate per stream) x 2 (for simultaneous record and playback) + (non-AV data)
The sizes of the read and write commands are matched with the host buffer sizes for each stream. Based on industry input, WD chose a medium-sized host buffer that correlates to 1024 sectors for write commands and 768 sectors for read commands. Non-sequential, non-AV data is injected in the write commands on a per stream basis.
The result of the test is a histogram showing what percentage of the read/write commands completed within a certain period of time. The test also displays the total drive idle time. According to WD, so long as idle time is greater than 20%, the drive passed for a given workload. Anything below that number cuts into any safety margin the system might have.