06/07/2024
Some thoughts on NVR4 RAID and drives.
There have been a few "what drives should I use in my NVR4" threads here from newer users. I thought I’d pull together some of my notes from field experience and debates over the past decade for those not aware of surveillance and storage nuances. The best approach for you depends quite a bit on your workflow. Like do you view recordings often, or just let them alone for the most part. And what kind of disk failure tolerance do you want. Or do you just want to load up the NVR with the max storage capacity for the least amount of money.
Industry has learned quite a bit about RAIDs and surveillance NVRs over the past decade. If you are going to RAID, RAID 1 and Raid 10 (Mirrored drives in a RAID 0 array) are the formats to use. RAID 5 is a actually a poor choice for surveillance NVRs. Statistically RAID 5 doesn’t buy you any protection from losing data for drives larger than say 4 TB (2TB max is more prudent) because there is a pretty good chance the rebuild will fail. If the system encounters an unrecoverable read error (URE) during the rebuild of a failed drive, the data is ultimately lost. UREs used to be an acceptable problem, but as drive capacity has increased, the odds of encountering a URE on a large video storage drive have increased to unacceptable levels. It is quite likely, that a RAID 5 array of large drives will fail to rebuild and therefore offer no protection against data loss. More on drives later but, for the NVR4 in critical surveillance applications, the best compromise is to use two large surveillance type drives in mirror (RAID 1), or an array of four enterprise 2TB drives in RAID 5. It would be a tad less likely to loose your video using mirror, the RAID 5 will be a tad better performer. If you feel lucky, push the RAID 5 by using 4TB drives. It would have been nice if the NVR4 engineers provided the option to use Raid 10, but they have recently said that is in the works.
There are drive failure scenarios that can not be accommodated by most RAID storage systems that are used for recording video. This issue is unique to surveillance video recording and seldom surfaces in RAID systems used by other applications. There are a couple nuances in using surveillance drives in a RAID 5, but the most significant difference for surveillance drives is error detection/correction. A typical drive does bit error detection and correction. It also verifies that what was suppose to be written actually got written and when an error is detected it marks the platter section as bad and writes the data to another area. There is a small performance penalty and on the rare occasions an verification error is detected, it will take a couple seconds to correct, during which time all writes (from all cameras) are suspended. The OS should gracefully handle this, but it may not have enough buffer to avoid loss of a couple seconds of video. Surveillance drives like the WD Purple avoid this by just not doing any bit error detection or verification.
One of the failure modes of hard drives is something called “Read Element Failure”. The drive is unable to read all or part of the data written to it. This could be the result of a complete failure of one of the read heads, or just a bad area of a disk that has not been relocated by the drive’s automatic systems.The trouble with surveillance drives is that disc errors can go undetected for quite some time. They are usually detected by the operator when they try to view an event's video but its not there. If and when the RAID system eventually detects a drive failure, it will attempt to rebuild the RAID set using the parity data recorded across all of the drives. If the RAID system also contains another drive that has a Read Element Failure, it is very possible that bad area contains parity data. If it does, the rebuild will fail and the RAID set is lost. This seems to be a once a year type occurrence.
So it appears that, if you can live with short rare hiccups (twice a year) use enterprise drives in the RAID 5. If not and can live with losing the entire Raid set once a year, use surveillance drives in the RAID 5.
If you are serious about preventing loss of video and avoid hiccups, use the RAID 1 mirror with surveillance type drives. Hopefully the NVR4 does some kind of health check to detect a bad drive. I hate it when your mirror RAID fails you because you didn’t detect the first drive failure before the second.
A couple less critical nuances of a choice between WD Purple, Red, and Seagate surveillance type drives:
- The RED’s overall performance will be superior to Purple drives due to access times. They also use more power than Purples. Some REDs use SMR which you want to avoid.
- The Purple uses ATA streaming and is optimized for a known read/write intensive workload. Workloads that stress the Purple up to a certain point are handled with great responsiveness. Once you exceed the drive's performance boundaries, less ideal results start cropping up.
- Seagate is optimized for pure write-intensive applications. Seagate’s surveillance write performance is typically 15-20% faster than WD's Purple, but suffers in read/write intensive environments. When the Seagate surveillance HDD is stressed, it responds more gracefully, kicking back fewer outliers. Seagate also includes RV sensors so it is more scalable than the Purples (an advantage in ~10+ disc arrays).
As with any statistical analysis, your mileage will vary, and the best is the worst enemy of good enough. For a low stress home application I would think these kinds of things are not as critical. When you have an important video clip, download it and store it somewhere with a proper backup.
RAID5 currently has a storage cost advantage for larger storage capacities. To get 6TB storage, you
need two 6TB drives in mirror or four 2TB drives in RAID5. Using quality drives, 6TB of mirror will be over $100 more than 6TB of RAID5.
CTTO.
