"It's also worth nothing that these tests were performed using benchmarking tools, not with actual cameras or even real video streams, so no real world performance issues were discussed"
So this is a "Surveillance Hard Drive Shoot-Out" without using surveillance cameras? OK.
IPVMU Certified | 06/20/14 08:05pm
In the article they used a utility provided by Western Digital that is supposedly designed to mimick surveillance camera throughput, with emphasis on large block writes and very little reads. The writer said they vetted the software to make sure they didn't try to skew results, but true they did not get actual cameras to do the testing.
The funny part of the article were the comments at the end with some in almost child like wonder going "ohhh, I want to put together a surveillance system now", and a number of posts speculating why SSD's might be a good consideration for surveillance.
Gee, an awful lot of the drive's specific workload is determined by the design of the recording software being used as well as the architecture of the recording system. Not all recording software lays down the video in the same way. Not all recording systems use drives the same way.
Small single drive or JBOD systems usually introduce other workloads on top of writing video to disk. An OS, a database for recording motion data or analytics data, etc. A multi-drive array is a different animal than is being simulated in this study as well.
We can all agree video surveillance drives are write intensive. Benchmarks that show good throughput are valuable. But I'm not sure this kind of study really tells me anything about how these drives are going to perform in my particular application.
Drive selection in commercial products tends to be
1. Does it have the overall throughput to meet my needs? (where back-of-the-envelope type numbers are a good start)
2. Does it have the right industrial characteristics (i.e, heat, mechanicals, power, etc)?
3. Does it have the right warranty?
4. Is it the right price?
Qualify, test (with your software!), monitor return rates, repeat.