Btw, this question / topic came up in the current IP Networking class. Keep the feedback coming in as this helps us prioritize what to cover.
Video Surveillance Storage Redundancy Statistics 2015
Published Oct 27, 2015 04:00 AM
100+ integrators answered "What percentage of your projects do you use redundancy or RAID for video surveillance storage? Why? What type?"
Compared to our 2011 survey, the percentage has increased significantly but a notable percentage of integrators still rarely use storage redundancy for video surveillance.
In this note, we break down the overall statistics, examine what types of storage redundancy is most often used and the key factors that drive selection (or not) of redundancy.
*****
*******, **** ** **** ** ************* ~60% ** ************ ***** ************ ********.
**** ** ** ******* ***** ******* 2012 **** **** ~**% ** ******** used ****.
*******, ***** */*** ** *********** ****** use ****, ***** ****** **** *** it **** ** *** ****.
Larger / ******* ********
*** ****** ******* ******* ****** * major *********** ******* ******* *** ****** camera ********. ***** ************ ** ******* camera ******** (~** ** ****) ****** used ********** (** *** ~**% *****). Hoever, ***** ************ ** ****** ****** projects (~*** ** ****) ****** ****** used ********** (** *** ~**% *****). In ****** ********, **** ******* **** minimization ***** ** ** **** ********* while *** ****** ** ****** ***** tends ** ** **** ******.
***** ********** ******** ******* ***** ** this *****:
- "*********, **'* ******** *** *** ******** > ** *******. ** ******* ********, it's * ******** ** *** ***** to *** ** ** *** (*****)."
- "* *** ******* ********** ** *** projects ****** *** ****** ***'* ***** only **** * ******** ** **** a ****** **** *****."
- "** *** **+ ****** ******, ** normally ****** ** *** **** ** all *****."
- "*** ******* **** ***** ******* ** cameras ** *** **** * ** these *****. ** ********* **** * would ** **** *** '******* ‘*************"
- "**** ** *** ******* *** ****** the *** *****. *** ******** ** these ******** ** *** *** **** or ********** *** ***** ** **** are *** ******* ******** ************."
- "** **** **** *****/****** ** ********** level ***** *** ****** ***** ** a ******* *** ******** *** * hard-drive ******* *** ******* ** * plus *** **** ********** *** *** User.
- "*** ***** ************* (*-* *******) ** either ** *** *** *** ********** or, ** ****, ****** *** ******. Perhaps */* ** *** ***** ******** use *********. *** **** ***** *********, the ******* ** *** ** ********* that *** **** ***** ******* *** footage **** ** * *** *****."
Reasons ***
********* *** ***** ** ************ *****, relative ** *** **** ** ****, was *** **** ****** ****** *** to *** ****:
- "******* ***** *** **** ** *********** by ******* *** ********** *** **** for ********** **** ******** **** *** for **** ****** **** ** *** as ********* **** ******** *****."
- "** ***'* **** ********* **** **** will ** *** *** **********."
- "******** *** *** ****** ** ******* cost *** ******* ** *** **** sensitive *** ** *******."
- "********* ****** ***** *** ***** **** to **** ***** **********."
- "** **** **** *** ******* *** **** interested ** ********* ** ********* **** redundancy."
- "**** ** *** ********* **** **** the *** ***** ***** **% ** the **** ****'* ***** **** *******. With **** ** ****, ***** ********* already **** ******** *** **** ** losing ******* ** * *** ****. Plus, **** ***'* **** ** *** the ***** ***** *** ****."
- "* **** ***** *** ********* ** they ***** **** *** ****** *** explained *** ******** *** **** *** not *** *** ***** ********* ** the ****.”
- "***** ** ****** ***** ****** *******, never ****** *******. **** **** ** strongly **** **, ****** ****** ** not **** ** ***** *** *****."
** ******, **** **** ** ********* about **** *********, *** *** **** smaller *****, *** ********** **** ******** evidence **** ** **** ** *** due ** * *********** ** (*) infrequent ******* *** (*) ********** ******.
Redundancy **** - *, *, **
***** *** *** **** ****** ****** but ***** *** * ****** ******, up *** ****** ** **** ****** than ** ****. ******, **** * number ** ******** ** ****** **** given:
- "**** ********** ***** **** * ** our ********** (***** **%) **** ****** sites **** * ***** **%"
- "* ***** **** ** *** ***** in **** *****. **** ***** **** up *****."
- "***** **% ** *** ****** ************ SYSTEMS. ***** **% ** *** ****** SURVEILLANCE *******."
- "* *** **** * ** ***** 2-disk ************* *** **** * *** bigger *************."
- "** ****** **** ** *** **** 6 ** **** * ** **** too *****."
- "** *** **** * *** **% of *** *****. *** ** **** requires ** ** ** * ***** crashes, *** ****** ***** *** *************. the ***** **% ** ***** *** enterprise ******* ** *** **** *"
- "**** ** ****** * **** ***** installed ** * ****** ******** ******"
- "**** ******* *****, ** ****** **** 10 ****** * ******* ** ***** drives *** ******* *****, ** * minimum ** * ****** ******* ***** (n+2) ***** * ** ** *** number ******* **** *... *** ******* online *******."
- "**** * ** ******* ******* ***** 30 *******. ****** ******** ** **** use **** ** ** *** "**** database" ****** *** **** * ** the "*******" ******."
- "**** * ** * *********** ******** for *** ******. **** * ********** replacing * **** ** ********* ********* by ********* **** **** ****** * really *****/******** ****** ** *****, **** often ** ********* **** '******' ******* staff *** ********"
- "***** **** * ** *. ****** we ******* ******* **** ******, *** customers **** ********* *** ******** ***** video ************ *** ***** *** **** coming **** ** **** *** **** their ******-***** ******** *** * ***** failure ** *** ********."
** * ***** ************, ***** ** the ***** ****** ** ** **** only * ********** ****** *** ******** drives. *******, *** ***** ******* **** RAID5 ** ****** * ****** ***** before ********** **** *** ***** ****** drive. ******, ******* ***** *** *********** as **** ****** *** ******, ********** this ****. ******* ** ****, **** users *** ******** *****, ***** ******** 2 ********** ******, ******* *** ****** cost ********. *******, ****** (*** *+*), everything **** *****, ******** ****** ****** and ******* ***** **** **** * and *, ** *** ******* ** requiring * ********** ***** *** ***** drive ****.
Comments (22)
JH
John Honovich
Oct 27, 2015
Mike Dotson
Oct 27, 2015John,
Another thing to ask about is HOW that RAID is defined.
For example...you have a larger number of HDD in the chassis... 16 to 40 (even 60).... How many HDD for the RAID sets get used.... say a 16 bay is created as two arrays of 8....or a single 16 way array with two partitions?
Mike Dotson
Oct 27, 2015FWIW..I asked our storage expert which way is the best.
Doing larger arrays then making partitions has a slight performance edge.
Also, 16 drives is an optimum spindle count for RAID5. More drives than that then go to RAID6.
Mike Dotson
Feb 25, 2016Playing devil's advocate on my own comment ;-).
Also consider the failing case when deciding how to make up the RAID.
If it is a large volume with multiple partitions... when a drive fails...it will affect ALL partitions.
If you make the volumes = partitions....then ONLY the partition with the failed drive is affected.
Since we all know that even Enterprise class HDDs WILL indeed fail...it may be best to give up the performance benefit and make multiple volumes.
U
Undisclosed #1
Oct 28, 2015I would be careful when using the term "redundancy". In the IP video space terms like redundancy, failover and high availablity are used often, in many cases incorrectly. This article speaks to RAID as a solution for storage "redundancy" however many times "redundancy" can be misconstrued as solution for a catastrophic server failure. Just a thought.
MI
Matt Ion
Oct 28, 2015It's almost an all-or-nothing with our customers. Pretty much all of the smaller/individual retailers just rely on the DVR/NVR's internal drives, and if a drive fails... oh well. Usually if they need to review an event, it's within a day or two of it happening; long-term events are things involving big bucks and more of a concern to the parent company or outside vendors (lotto, credit cards, etc.), and if the video is gone, it's someone else's problem.
The only customer that does use RAID storage, it's still on less than half their sites, but for almost all sites that have it, we're using RAID6 with a hot spare. There was actually one case where we had a drive fail in a RAID5 array, and then another fail before the array had finished rebuilding on a replacement drive, resulting in the loss of everything. After that the decision was made to go to two-disk redundancy, and since it's not always possible for someone to get to site right away to replace a bad drive, the hot spare was decided on as well. So far there have been no data losses on those systems.
Rob Hammond
Oct 29, 2015This thread is helpful to us on the design side, understand what you guys on the install side are experiencing!
To reiterate what 1 said above, RAID is not redundancy, it is resiliency, true redundancy would be a separate server mirroring the active server; on active server fail, auto switch to the backup server - I have never heard of this type of deployment in video surveillance.
Another redundancy solution is to store locally when the server goes down, the camera detects a server failure and it stores locally, until the server comes back on line. My question is how well does this work in the real world, we have SD cards in most cameras now, have people been deploying this as fall back? Does it work as advertised?
MI
Matt Ion
Oct 29, 2015I did set up a site once with edge recording in addition to central NVR... the cameras simply recorded all the time, not only when the server was offline. This was done because the previous DVR had been stolen along with a number of cameras.
Luis Carmona
Oct 29, 2015Sorry I have to disagree with you Rob. Granted, it's a matter of symantics, but RAID is redundancy. Redundancy is in the very name- Redundant Array of Independent Disks. Just because it is not an entire server being redundant is not a reason I think to not call it "redundant".
What I often fight against, and I think much more important than getting people to consider RAID either redundancy or resiliency, is to get a person to not call RAID a backup. On a strictly technical definition, you could say RAID is a "backup" because there is a backup of the data bits on more than 1 disk, but the common implication is backup means a backup copy of files, such as a database file if it should become corrupted, which of course RAID is not. RAID is not a backup of files. But it is a redunancy of data bits.
*All comments above exlcudes RAID0.
TS
Tariq Saleh
Nov 04, 2015Hello,
RAID 6 is better than RAID 5 in terms of redundancy. However, the RAID 6 has write performance penalty due to overhead associated with the parity calculations. RAID 6 performance varies depending on the manufacturer’s storage architecture.
Thanks
AN
Alan Nagl
Feb 23, 2016This conversation is really interesting to me. During the past couple of years, few of my/our customers report using RAID 5 or higher, and now this poll shows it has grown to >50%.
As a hard drive guy, I tend to think of RAID as an insurance policy. It can provide partial protection against data loss, but it comes at a price, and it is not 100% protection.
Also, it's important to remember that RAID5 and higher is actually a tougher environment for drives. Given the same model/same capacity, drives running a RAID5 or higher file system will have a slightly higher failure rate...
JH
John Honovich
Feb 23, 2016AN
Alan Nagl
Feb 23, 2016Well, certainly it does vary, depending upon their segment of the market and their tolerance for cost, but for us the WD Purple series is by far the favorite.
We also carry the Purple NV, which is a better choice for RAID5 and higher file systems, and of course, the Enterprise class drives are picking up steam as the larger NVRs come into their own.
I'd say we're seeing 90% traditional streaming drives (Purple), 5% entry level RAID (Purple NV) and 5% Enterprise class (WD Re).
U
Undisclosed #2
Feb 23, 2016JH
John Honovich
Feb 23, 2016AN
Alan Nagl
Feb 23, 2016Yes, many of our customers are in the lower end of the market and just don't have room in their customer's budget for redundant storage.
JH
John Honovich
Feb 23, 2016I believe that. If we were to break the numbers out for segment / size of system, you'd likely see <10% for the small / budget market and >80% for the 'mega' market.
AN
Alan Nagl
Feb 23, 2016Absolutely. And as the volume of your data increases, so does your liability/value of said data.
Also to be considered is the cost percentage effect. If you're only storing a TB, redundancy may be a 35% adder. For a Petabyte, the RAID 6 adder may be only a few percentage points.
AN
Alan Nagl
Feb 23, 2016More RAID 0/1, then RAID 5. 6 seems rare, but I'm not sure why.
Luis Carmona
Feb 24, 2016One aspect of RAID I think some people miss and need to be careful about, is where is the RAID parity written or stored? I remember there were some lesser expensive RAID controllers that stored parity info in the RAID controller. That meant if the controller went bad, you lost all your disk info. Most however write the parity info on the drive. That means if the controller goes bad, or you want to move the disks to another chassis, you can do so and the RAID info can be read by the new controller and all data will be there and accessible.
It's been awhile since I've researched and worked with these nuances of RAID controllers, and maybe that's not the case anymore in some controllers that RAID info is only in the controller. But anyone using RAID should ask the question, "What happens if my controller fails?"
U
Undisclosed #2
Feb 24, 2016I remember there were some lesser expensive RAID controllers that stored parity info in the RAID controller.
I've never heard of that before. Where on the controller did it store it?
Wouldn't they be more expensive if they did store it?
Luis Carmona
Feb 24, 2016"I've never heard of that before. Where on the controller did it store it?"
I don't know, maybe NVRAM or something? It was an Adaptec controller and I learned some of their models stored RAID information (maybe I shouldn't have said parity information) on the controller and some models stored RAID info on the disks when I had to call their tech support once. But that was about 7 years ago.
My main point was, make sure you have the method for recover for the loss of your controller along with the loss of a disk planned. Get to know what's involved beforehand. Heck, you might make a sale providing extra parts. If you don't stock them yourself, would it be prudent if the customer had some...?