The WD Surveillance Benchmark also displays histograms of the read and write completion times. When you start to stress the drives, writes take priority over the read operations, and their completion times start to drift upwards. This is what you want to see; if you have to drop data, you'd rather lose information played back than the stream you're trying to record. For all drives, write completions were always less than 20 ms.

In the graph above, we are looking solely at read completions times.
In the first test case, all products perform well. Everything except the WD Red has 50% of its read commands complete in less than 20 ms. Moving to the more demanding test cases proves to be problematic for the non-enterprise drives.
The Seagate Surveillance HDD has slightly better read distributions than WD's Purple, and in the next section, we'll see how the actual distribution looks for both disks.
Looking at the enterprise drives, you can see that they are better equipped to handle more demanding workloads. In fact, the Seagate Savvio 15K.3 just plows through the tests. Even though the idle time for the 32/720p/30 FPS lands at 9%, almost 100% of the writes and reads are still under 20 ms.
- Introducing Surveillance-Specific Hard Drives
- Western Digital Purple
- Seagate Surveillance HDD
- How We Test Surveillence HDDs
- Results: Transfer Rate And Access Time
- Results: WD Surveillance Benchmark, Idle Time
- Results: WD Surveillance Benchmark, Read Completion
- Results: WD Surveillance Benchmark, Distributions
- Pricing
- High-Capacity Hard Drives Built With Surveillance In Mind
Are there any demonstrable effects on performance of having these drives in a small external raid array (of perhaps 4 drives)?
If you're going to wander into the USD1/GB+ territory, even just for informational purposes, please include an SSD in this mix to be fair. SSD performance/price just might validate people buying them for surveillance drives.
Thank you.
Indeed, or any typical Enterprise SATA model (Hitachi HUS, Seagate ES2/NS, etc.)
Come to think of it, given the consequences of not being able to identify a suspect or
obtain other relevant visual information due to dropped data, as AndrewJacksonZA
says it would be interesting to know how these drives compare to various high-capacity
SSHDs/SSDs, eg. the Seagate 4TB ST4000DX001, Samsung 840 EVO 1TB (which includes
AES), and (high-density option, power saving) the Samsung MZ-MTE1T0BW 1TB mSATA.
The higher cost/GB of these products is surely more than worth it given the intended task.
Ian.
Article idea: No offence to the writers at Tom's but this is the first interesting article I have read in a long time. Could we get some more articles like this? Maybe some articles comparing onboard Intel RAID with different popular card options? Or comparing how different drives perform in different RAID configurations and workloads? I get the feeling that these drives would perform quite differently as they are really made to work as a team rather than as solo drives.
You would need one grossly under-engineered surveillance system for this to really be a problem since there is nothing happening 99% of the time and those drops would need to conveniently happen during the 1% of the time where you need data and your system administrator would need to have somehow failed to notice and fix the issues in-between events if they were so bad as to render the system unusable for its intended purpose.
Most of the time though, surveillance recordings are merely a nice convenience in case something goes wrong but are not considered critical outside of casinos, banks and few other (very) high security applications that have their own IT departments or dedicated vendors working on their video archival needs and are unlikely to take their hardware recommendations from enthusiast sites like THG.
I doubt any normal company would waste SSDs or 10k/12k/15k RPM HDDs on video surveillance storage. They would be more likely to use standard HDDs like WD Black / Red / Green / Blue.
obtain other relevant visual information due to dropped data, as AndrewJacksonZA
says it would be interesting to know how these drives compare to various high-capacity
SSHDs/SSDs, eg. the Seagate 4TB ST4000DX001, Samsung 840 EVO 1TB (which includes
AES), and (high-density option, power saving) the Samsung MZ-MTE1T0BW 1TB mSATA.
The higher cost/GB of these products is surely more than worth it given the intended task.
Ian.
I really debated on whether to include SSDs in the evaluation. The problem is that because of $/GB and write endurance, it would only make sense to use them on a smaller scale setup, which is where their benefits (simultaneous high speed IO), are greatly reduced. Also, their performance would skew the graphs to the point where it would be hard to interpret the results of the HDDs.
I would have like that also, but we only had access to so many drives, most of them SMB/enterprise.
Have you seen that article on some other site which is testing SSDs to destruction? Going for 1PB+.
In reality they perform much better than most people might assume, and that's just consumer SSDs.
Enterprise SSDs wouldn't have that issue. As for cost, as I say given the intended task, I really doubt
the cost would that much of an issue - this kind of hw setup is similar in importance to a backup
solution: costly, but well worth the investment given the potential costs in the event of a major incident.
Plus, the absolute costs of the drives isn't that much compared to all the other hw required, and the
power savings will add up over time.
hate to say it but given where the industry is going, including an SSD or two in the mix, both top-end
consumer & pro, is exactly the data that relevant users really need. It's similar to the benefits for pro
users of using pro cards: the advantages are usually worth the much higher GPU prices.
Also note I mentioned hybrid drives aswell, not just SSDs.
Ian.
Hybrid drives would be pointless for video surveillance since most of the footage is write-once-never-read.
http://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte
It really surprised me!
Having more cameras may increase the total performance requirements but it also increases the total storage space requirements. At 1GB/h (2.2Mbps average) per camera on a 16 cameras setup, you already need around 400GB/day so you already need one $250 500GB SSD per day instead of one 3TB HDD per week.
So, if you want to keep your footage for at least a week, getting one 4TB HDD for ~$200 per 16 cameras to simultaneously meet both storage and performance requirements makes a lot more sense to me than spending around $2000 on a 3TB SSD.
That doesn't take all factors into account. If an SSD solution means much more reliable capture then the extra
cost may well be regarded as a worthy investment.
Ian.
I think it was one of your posts elsewhere that had the link I originally followed, but thanks again anyway.
Ian.