Is Storage So Inexpensive You Should Record Continuously Now?

Statistics show most use motion triggered recording.

However, one member objected, with the following explanation:

"$25 of storage for a month of continuous recording on a 1 Mbps average camera (average interior camera installed is normally $800-$1000 and exterior is $1200-1800) or multiples of this for higher resolution or bandwidth cameras. These days, motion detection is used for event tags, and using on motion should only be used for close views like doors."

What do you think?

Vote:


A 1Mb/s stream recorded continuously for 1 month requires 325GB of storage.

If you have 16 of them, that's 5.2TB.

If you recorded motion only, depending on camera positioning / scene complexity, you'd probably be looking at 2 - 3TB instead of 5.2TB. The argument then goes, you are only eliminating a single hard drive, which is not much money compared to 16 cameras.

Of course, 1Mb/s is a low estimate for an HD camera bitrate, presuming ~10fps and not a terribly complex scene (also no huge night time bandwidth spikes).

Screening hours of unnecessary video for an incident is painstaking. Recording based upon motion and other triggers reduces a time-consuming job considerably, making it more likely the surveillance system will be used to maximum advantage.

Why screen hours of unnecessary video? Why not just screen the motion-detection-flagged video? This would seem to require about the same amount of effort as it would take to review video that was only recorded when motion was detected.

Continuous recording doesn't seem that much more expensive, but certainly could eliminate confusion occasioned by brief snapshots of ... what happened (or didn't happen there)? ... kinds of motion-activated video. Also, some motion detection algorithms don't store the previous several seconds of buffered video, which can sometimes miss the initial event.

We had a case several years back in which we had to prove something did not happen. A person claimed to have frequented our dining establishment and become violently ill. In fact, that person (with a history of ill will for our business) had never darkened our door or consumed our food. After passing the surprise health check with flying colors, we recovered 100% video coverage of 8 cameras for the entire day, proving our assertion that the individual had made false allegations in order to leverage the coercive power of the state to unjustly bludgeon our business. Had we recorded only on motion, with significant gaps, this "proof" would not have been nearly as compelling.

Similarly, not so long ago IPVM ran an article that appeared to indicate that a courthouse's record-on-motion video was so unreliable that a judge disallowed it as evidence.

I suppose your storage strategy depends upon your objectives. Based upon our experience, and the apparent experience in this courthouse incident, what value would you place upon unambiguous continuous recording?

Looks like that is a comment I had on another thread. My comment about the motion flag was based on searching. Still, exterior cameras have to always be run continuous especially since with high definition we are usually looking at larger scenes. The thing is, the goal of our industry is to provide forensics for evidence and the percent chance of obtaining goes up exponentially if you record continuous. So what if it takes more time to search. 2 tb certified drives now are 150 bucks. I normally use vbr recording with reduced agc settings with a goal of 1 Mbps average on a 1-2 mp camera (main-high profile). This is achievable (reducing agc even a little bit helps alot to reduce bandwidth below trade show settings). Many integrators also think frame rate is linear in regards to bandwidth. Average cameras set for 1 fps continuous with higher rates for motion use more than half of the storage compared to continuous when you take into account the size of the key frame related to prediction frames. My main comment is that if a customer spends 1000 dollars on a camera, for an extra 15-20 bucks, they can increase their chance quite a bit for usable frames by going continuous. The bandwidth for the camera is used no matter if you store or not.

Jeffrey, Thanks for starting this off. Good explanation above.

Two things:

  • As you know, but for others, turning down gain almost always results in a darker image in low light, which is frequently a problem of its own. See: Testing: Gain / AGC Impact on Surveillance Video. Because of that, we recommend using VBR with a cap to suppress worthless bandwidth spikes in low light.
  • There are definitely people spending $1,000 per camera on average, but I think the industry average is closer to $400 for fixed IP cameras. As the cost of the camera gets lower, the more cost conscious the customer often becomes overall.

"Many integrators also think frame rate is linear in regards to bandwidth. Average cameras set for 1 fps continuous with higher rates for motion use more than half of the storage compared to continuous when you take into account the size of the key frame related to prediction frames."

Jeffrey - Are you saying - for you example - recording continuous would require less bandwidth? Is this if you were recording at 1fps continuously the entire time? Do you know of any articles you can link to that can explain this principle? As an end user I understand I & P frames just fine, but have never considered the the point you made. I'd like to understand it better.

Tony, see our report: Testing Bandwidth vs Frame Rate

For instance, if you increased frame rate from 5fps to 15fps, bandwidth almost certainly would not increase 3x (i.e., linearly). It would be closer to 1x or 2x increase (less than linear).

Conversely, dropping to 1fps does not give you linear bandwidth savings. The reason is that you still need I frames every so often (usually once a second) and those are far larger than p frames. See: Test: H.264 I vs P Frame Impact

400 dollars? No wonder I only do enterprise solutions. On the low cost end:

250 bucks for camera plus markup = 375

Cabling, terminations, and hardware = 30 plus markup = 50

Camera license= 150 bucks

Installation (4 hours in, 6 hours out) = 325 for interior

General training, programming, and support (average per camera, 15+ system) = 100 bucks

Camera % of headend= 200 bucks

Networking % of project = 25 bucks

Total = 1000 bucks per camera (on average)

How do you come up with 400 bucks? (in my experience, 1000 x inside plus 1800 x exterior = general system price).

I did just sell a couple 4 camera hikvision systems parts only for 1600 each, so I guess......

I meant on a camera, not the whole installed cost.

That said, your solution cost, I think, is still higher than average. We see a lot of people using entry level VMS licenses (at $50 - $75 per channel) or NVR appliances that are even less expensive.

That said, the 'average' install is not an enterprise one, by definition.

Hey, you may have spawned another thread. Most of my solutions ate based on Exacq Pro or Milestone Pro. I believe most discussions I see on this forum are based on the same and also include Avigilon, DW, Genetec, etc which are similar priced. Even by reducing licensing by 100 bucks, you can not get far away from an 800 dollar per camera (budget) cost with an 1100 true wdr cost and 1500 exterior cost unless you price labor at minimum wage. I have to assume that it is manufacturers and enterprise level integrators that comprise a large percent of your reader base in addition to those entry level professionals gobbling up information from the hall of famers. (I miss the guy that only piped in to critique grammar and had nothing to say about the industry. Always good for a laugh.).

Jeffrey, yes, we do have a lot of larger scale integrators and they do spend / charge a lot on cameras.

On the other hand, many of these projects demand higher frame rates, higher resolution, longer storage duration, etc. so the storage costs can really add up.

Basically, I think 1Mb/s per camera is on the low end for enterprise systems. I more commonly see in the range of 4Mb/s.

I am not saying there is a right or wrong bandwidth level. My point is that for people with higher average bandwidth per camera, the per unit storage costs become more impactful.

I think, whenever cost allows, video should always be recorded full time, and motion searching done retrospectively (assuming the video system has that feature). I have had cases where proving something didn't happen, for legal reasons, is as important as proving something did. With motion only recording, the court (and even myself) can never be 100% sure that the event didn't happen as it's too easy to misconfigure sensitivity or zones or due to some disk or processing error the motion event was never triggered.

Also typical motion recording settings may be designed to ignore objects far away on the grounds that faces can't be recognised at that distance anyway. But for serious incidents this video is still useful as it shows how many people were involved, which direction they came from and where they went to. It may also help determine which other cameras may have got a better image.

It is all about bugdet though ;)

I have to agree that full time recording is the ideal way to go considering the many reasons listed in the posts above. As it seems, most people are concerned with the cost of storage or storage space required due to the use of megapixel cameras. At most of our facilities where we are using either 2MP, 3MP or 5MP cameras for outdoor surviellance we have opted to record continuously as we cannot afford to miss any recordings due to the fact that motion may not be detected for whatever reason. We also initially were incurring a lot of motion caused by trees, etc. Yes, you can select motion zones or mask off areas which are not relevant but there are many variables when dealing with trees, wind, weather, etc.

Initially we didn't think we could afford 30TBs, 45TBs or even 60TB of storage to allow for 30 day retention periods of many megapixel cameras recording continuously. After several RAID failures of storage systems which were about 8TB or 12TB max in size due to the expensive nature of RAID Storage Solutions we considered using Veracity's COLDSTORE Solution. We were able to get up to 90TB of storage (using two COLDSTORE Units) at three of our locations and record continuously for 30 days. If I would have tried to find a RAID Solution with 90TB of storage space it would not have been budget friendly. With the COLDSTORE Solution not only did we save on the actual hardware costs, but we also save on electricity. Each COLDSTORE unit only uses a max of 42 watts on a single PSU. The failure rate of the harddrives (off the shelf) dropped drastically since 80% of the time only two hardrives are on. Less coolling needed in our server rooms was an added benefit as well.

Here is an IPVM Review on the COLDSTORE Unit:
http://ipvm.com/updates/1509

Here is another discussion on IPVM which talks about 50TB+ Storage Systems:
http://ipvm.com/forums/forums/video-surveillance/topics/what-are-you-using-for-50tb-surveillance-system-storage

The snapshot below shows a comparison of Energy Savings:

In my personal opinion and experience the COLDSTORE Solution has allowed us to make continuous recording of cameras affordable and easy to maintain.

OK, I will bite on this, which may cost me additional posts to back up my claims.

I know John stated that the more standard was 4 Mbps, but I do not buy this. I am a firm believer in VBR over CBR. Why on earth would you want worse video when action occurs. And why would you want top quality video when nothing occurs. Does not make sense. I took so long on this because I had to hijack a customers system (via Logmein) to confirm my results. (And had kids/camp situations during vacation week)

First of all, I live in Maine, where we have over 12 hours of night in the winter. I always design for "night" when I do systems. In my experience, we are in a mode of "bigger is better" but this is absurd. I have tested 1,2,3, and 5 MP cameras at night and none provide any better that a camera with a 1-2 MP sensor, so I stick to 1-2 MP (or 720-1080 if lower installed elevation). A matter of fact, I always though the Arecont AV3135 camera was "genius" with a day 3 MP and night 1.3 with no cut filter (Do not use now, but the concept is good). Anyway...

I went to a customer site and tested my "VBR" cameras. (Like I said, I do not understand the CBR philosophy). I was looking at "idle state" which occurs on average 80-90 percent of the time. All of the 1 to 1.3 MP cameras (1/30 or faster shutter) run at no greater than 600 kbps on main profile cameras in idle state (I do use High Profile cameras, which show better results). On total motion scenes this can jump to 2 Mbps, but on average, I would say 1 mbps (wait, is not that what I said in my post?). I decided to vary fps to see increases in bw/storage to back my claim that higher frame rate is not linear (do not direct this reponse to a past IPVM test). On a 1280 x 800 camera with multiiimage WDR, in idle state, and Q of about 30, Gov 1 sec, I was getting 450-600 kbps h.264 no matter the frame rate with a 1 fps value of 50 kB H.264 and JPEG of 85 kB. I verified this by sabataging the customers 100 camera system, setting 4 or the 1 MP cameras (VBR) to 10 fps, and only recording these to continous (none of the rest were recording). This was after hours, so there was no motion. The average per camera with mid level compression was 500 kpbs (I looked at the size of the storage file for the hour and determined average Mbps and then "un-hijacked' customer system).

Could someone tell me why they are running 4 Mbps BW on cameras. Is this for CBR mode just to cover the 10% instance when motion and complex scenes exist. I guess I can see a 5 MP camera in an airport or casino (no wait, Carl said he needs 30 fps so this can not be so), but how may interior locations need high resolution cameras. (I work daily telling customers why the 20 MP surround camera is a day only solution (competitors without Electical Engineering degrees like I have make these recommendations) and why the 5 MP fisheye can not view an area effectively unless in a 24 x 7 mall.

To restate, my theory is no greater than 2 MP for exterior (use more if needed instead of higher resolution) and normally 1.3 MP 4:3 for halls or low hanging 720/1080 with 16:9 for larger areas. Depth of view reduces ability of higher resolution cameras in hallway areas, so < 1.3 MP with more cameras (heck, 1.3 MP domes are now <$150) is the norm. Strive for 1 MBps average (wait, once again, is not that what I said in my original post). Use multiimage WDR for entry doors and exterior window/glass areas. Always use < 1/30 second shutter. Get rid of 2x and 4x Sensup. (trade show marketing, I think).

1 Mbps ( I say less than this per average per average camera) = 325 GB per month. 2 TB drives = $150 continous per camera @ 30 days = $25 (continuous motion cost per camerea per month)

See original post. (If anyone backs up the claim by John that an IP camera can be installed, programed, and licensed for $400 for a system, I would like to hear it - our average is $1000 per interior camera installed, licensed programmed, and commisioned - if they can, I was thinking of getting into solar energy).

"I know John stated that the more standard was 4 Mbps, but I do not buy this."

No, I did not say. This is what I said:

"I more commonly see in the range of 4Mb/s."

I am not sure why you don't 'buy' this. Are you implying people are lying to me?

You continue with:

"I am a firm believer in VBR over CBR. Why on earth would you want worse video when action occurs."

I think your point is that you do not agree that 4Mb/s CBR should be used. Nor do I for that matter. I am saying, though, as an empirical matter, lots of sites use 4Mb/s CBR or a 4MB/s VBR 'target' for HD video surveillance systems.

Additionally, for 'enterprise' deployments, many sites use 15fps or 30fps (even if they are 'wrong' or 'imprudent' to do so) and this drives up bandwidth requirements as well.

Finally, lots of cameras, using VBR, have very significant (5x to 10x) bandwidth sites at night compared to day.

Finally, lots of cameras, using VBR, have very significant (5x to 10x) bandwidth sites at night compared to day.

Would you change from VBR to CBR (or lower the VBR cap) automatically on Day/Night transition if it was an actual option offered?

I'd just use VBR with a cap 24/7. The practical reality is that night time bandwidth surges tend to be worse than anything one experiences during the day (in terms of bandwidth spikes).

VMD is not merely used as a feature to reduce recording space but as a means to only recording relevant information. So when retriveing archive recordings, it is a lot easier to viewing movement as to no movment. Why would it be necessary to record information when there is no motion other than the incidence of a guard sleeping (which will be viewed at inset and outset of occurrence)? In the South African market, the prescription is towards VMD recording as opposed to continual recording.

Why would it be necessary to record information when there is no motion other than the incidence of a guard sleeping.

Here is a reason: Outdoor scenes a notoriously hard to set because of lighting, wind, sprinklers, leaves and wildlife to name a few. Have to come back at night (oooh) and maybe even during a windy night. Takes a lotta time=money. And the first time (or second) you miss something that could have helped you could cost you too. Money+money=more money.

So I guess the answer is: How much is your time worth? If storage was free would you still only record on motion? Also, modern VMSes create events at the same motion threshold whether they record only on motion or continously. Searching thru/skipping events is just as easy on VMSes I've used, though perhaps exporting clips could take longer.

Here's another reason: During post-mortem when you are 'following' an unknown peripatetic moving between many areas in a structure, there are often slight clues not found on the primary cameras, far to slight to trigger VMD, like shadows, that let you piece together someones path. Plus it makes the roll of multi-view playback far more seamless and understandable.

As for the 'guard sleeping' concern, may I suggest loading up some on 'Snooze Analytics".

I prefer to keep constant recording in addition to motion recording. I have recently been setting up my analog installs with D1@7FPS normal recording (constant) with VBR cap @512Kbps, then also motion recording with D1@7FPS with a VBR cap of 1024Kbps. While the normal recording has the same resolution and frame rate, the lower VBR cap cuts down on the additional storage needed. To be honest, most of these analog installs are small camera counts (8 cams usually) and a single 2TB drive is more than enough for a full 30 days retention. And the best part is my customer still receives motion cues that they can use to cut down on forensics time.

Now for our IP projects, they tend to be larger camera counts and vary greatly depending on the type of customer. For instance, an apartment complex we are working with right now in the planning stage is going to need 15 IP cams (13-1080p outdoor, 2-720p indoor) and we are proposing 9TB (3x3TB) of disk space. At $450 cost of the actual drives, how could we not think of drive cost as a minimal part of the cost of the project. It is a mere fraction of the overall cost.

We will be recommending constant recording for almost all customers going forward, unless Thailand floods again. (too soon?)