200 Multi-Megapixel Cameras Storage Requirement?

I have a client who wants 200 multi-megapixel cameras, continuous recording, minimum of 30 days recording, full frame. 150 of the cameras will be 5MP and the other 50 will be 10MP.

How much storage will I need? What type of storage system should I use?

Make and model of the cameras would help slightly... assuming Areconts, the 10MP can only run at 6fps, or maybe its someone elses cameras.

Ok, I used the AV iOS app to run some calculations. TL/DR: Holy shit! It's 2 - 3 PBs

Here's the details:

  • For a 5MP H.264 stream at 13.5fps, 100% motion, the average bandwidth was 6.4Mb/s (sounds reasonable). Total for 30 days per camera was 12.5 TB
  • For a 10MPH.264 stream at 6.5fps, 100% motion, the average bandwidth was 8.1Mb/s (sounds reasonable). total for 30 days per camera was 15.7TB
  • With 150 5MP at 12.5TB each, that's 1.875 PB; With 50 10MP at 15.7TB each, that's 785TB.
  • All total - ~2.6PBs / 2,700TBs

Let's say the average hard drive used is 2-3TB, that's ~1000 hard drives (not including what's needed for redundancy/RAID, etc.).

Ways to reduce:

  • Limit the number of days stored (drop to 15, cut by 50%)
  • Use motion only (say only 50% active)
  • The two of those combined drop you to 650TB, still very large

Fun fact: These 200 cameras replace exactly 3750 analog cameras :)

Anybody have thoughts or suggestions?

"Fun fact: These 200 cameras replace exactly 3750 analog cameras :)"

More, if they're all 180-degree or 360-degree cameras...

>Ha! ha! ha! Math is fun...the Avigilon way!

"Fun fact: These 200 cameras replace exactly 3750 analog cameras :)"


More, if they're all 180-degree or 360-degree cameras...

Other ways to reduce:

Limit frame rate where possible (some scenes may only need 1-2fps, some may be fine with 10-15fps, etc. - there's probably no reason ALL cameras need to run full-out).

More careful motion masking (block out things like windows, streets in the backgrounds, monitors or flashing signs in the field of view, etc.) - extraneous things that will trigger unnecessary recording.

If your network is solid, you could try increasing i-frame interval.

Use VBR, so bandwidth will drop with less contrasty/complex scenes and less motion.

If client insists on constant record on some or all channels, try setting a low constant framerate and kick to higher rates on motion (assuming cameras and/or VMS support this).

Of course, all suggestions above (yours included) rely on convincing the client to go with what's *necessary* rather than what they think they want...

Matt, Arecont defaults to VBR, so their number is most likely an average of VBR over the course of a day. With Arecont VBR, I'd be especially concerned to measure and verify low light / night time bandwidth consumption which could easily spike to over 15Mb/s with those resolutions.

True, it will also depend (somewhat) on just low the light will be in the target areas. And on how the cameras handle it. If the camera can switch to B&W output at night, it will probably do a lot better than a color camera under the same conditions/

Very obvious but it has not been mentioned: go for lower resolution cameras? Isnt a 2mp camera almost sufficient for all applications - I know this is a naive question but maybe the OP has to explain why they need 5mp and 10mp cameras in the first place?

The way I see it, the OP has two choices here: bid on providing sufficient storage for the project or convince the client they don't need such high pixel counts/frame rates. That would obviously depend on the application and the expectations of the client.

I'm a firm believer in "Just Enough" surveillance. While it requires more knowledge of requirements and more careful planning, in the end it supplies the benefit of lower cost.

I like to determine optimum camera placements and required FOV and choose a camera and lens combination that, as close as possible, fits that need. For instance, we've had a number of entities recommend 2MP (and more) cameras for our table games. Our tests have shown that XGA is not quite sufficient while 1280x960 or 720P (depending on the camera's capabilities) is more than enough.

We are choosing one of the two, rather than higher pixel count cameras. Higher resolutions would only require greater bandwidth and storage while providing no substantial improvement in our ability to see what we need to see. YMMV.

I vaguely remember a slide from one of the camera manufacturer where they can perform a bit of a trick feature which will give the focused area higher fps while the rest of the frame is way low like 1 or 2 fps, is this something any of you have seen in the market yet or it's still a R&D target? Suppose such feature is available then it will be a big help for this case or not? For that many cameras looking at the same premise it would bound to have lots of waste of unimportant scenes, will be a different story for a city surveillance I imagine. That's my two cents.

How many of 5/10 MP cameras would be indoors? It would be easier to control the bandwidth at night for the indoor cameras.

I am pretty sure after looking at the storage requirements the client would be easily convinced to reduce FPS.

Matt, switching to b/w mode helps modestly but, in our testing, even in b/w, VBR cameras in dark scenes tend to have severe bandwidth spikes.

Heng, Panasonic offers VIQS allows defining an area of higher vs lower desired quality. It does not significant reduce bandwidth consumption nor does it make a big difference in quality. It's a nice 'spec' item but not much practical value.

One technique that could help is to use privacy masks on areas that are not of interest. This could reduce bandwidth substantially. See our test results on reducing bandwidth with privacy masks that includes tests of Axis, Bosch, Panasonic, Sony and Arecont.

I would want to know why the customer wants 5 & 10MP cameras.

Is it because of our American thought process, Bigger is better, more is better??

What is the customers level of knowledge on this subject?

Other than that after reading the comments above my approach would be like Carl's.

Select product based on the application, scene, indoor? outdoor?, Day/night? VBR? All the variables must be taken into consideration for the respedtive scene. All to meets the customer requirements.

Question: is the customer going to pay for the time it takes too put a system of this caliber together?

A system of this size would take considerable to put the cost together.

Luis and others are right, this customer probably needs his expectations managed a bit.

One of our larger clients, when we started with them, we set up one of their sites with a few 1.3MP cameras, and they loved the HD picture... over time we went to 2MP, 3MP, and even 5MP cameras in several instances. At the end of the day, they found the 5MP models just didn't have the low-light performance they needed, and honestly didn't seem to provide a lot of benefit from the increased resolution (large highly-detailed areas of complete darkness is rather pointless).

On their latest site, with 42+ cameras, a dozen of them are Axis P3384s, which are only 1MP, but which have fantastic WDR and low-light response, which is what's really important in this case, and frankly, the results are far better than any of the 3MP and 5MP cameras we've used for them thus far.


Umm... I don't know why anyone else here hasn't stated this but if you are going to implement and support your customer you should be the one calculating these values based on your design. This isn't something you want to experiment with or a "get your feet wet" project. There are so many more implications / considerations here other than storage all of which you will need to know your bandwidth requirements - network & processing just to name a couple. If you don't know what type of bandwidth and storage requirements based on the solution you will provide your customer you are in deep doo doo.

Your customer may need the cameras stated and the associated 1-4 petabytes of storage but it appears storage is really only a small part of a pool of undisclosed questions.

I think before I did any calculations I would want to vet the client. See my comment above.


I work for Spectra Logic and we offer storage solutions for Video Surveillance. I think I have some value to add to the conversation and its purely for educational purposes.

Let's just assume the customer needs the multi-PB storage solution. I know the VS industry is use to 100% disk solutions and is very reluctant to do anything but that. I 100% agree it makes sense in most installations. However, when you get into the top echelon of customers who play in the PB plus range, why would one not compliment disk with digital data tape? The TCO becomes very compelling in that PB plus range. Disk is needed for fast retrieval in the first week or so. But often the archive is never viewed again. So why not put it on a more cost effective tier of storage that requires little to no energy in the archive? What this does is it frees up the budget from storage to more pressing areas of the overall solution. Who wants to spend all the money on storage? All one asks is, can you wait two minutes for the content to come back, in trade off for greater than 50% acquisiton cost savings. Obviously you will also have OPEX savings too.

Undisclosed "Umm... I don't know why anyone else here hasn't stated this but if you are going to implement and support your customer you should be the one calculating these values based on your design."

Agreed. If you have a job that size but you have to ask those questions, I really think you need to take some of John's classes or something similar beforehand.

He's in the class actually. This came from the class. I was hoping to open this up to get other's perspectives, not have him be attacked.

Tape? LOL

All I can say is, we went digital in 2003 specifically to get away from tape. Since then, we've had a few vendors approach us about using RAID to store streaming video for 3+ days and tape for the rest. I'm sorry, I wouldn't trust tape of any format.

Back to cleaning heads and drive mechanisms and dealing with crinkled tapes? NEVER!!!

@Carl. I have seen your responses before and know you will never be convinced. But rest assure we are not talking your VCR tape. The largest archives in the world reside on digital data tape and it is more secure than disk. Reliability, you'd be amazed. All the most demanding big data enviroments in the world are using it because it is really the only option. Cost savings, nothing compares. It is being used in VS today in these high end environments. I remember Jan Berker last year saying he was installing tape in high end environments left and right, and you mentioned why and doubted it. It helps customers, get what they want. It really makes sense. Tape has been around since the 50's and is incredibly reliable and robust. And with more than 80% of the world's content residing on digital tape why should Video Surveillance not consider this? I respect your position, but respect mine too cause the customers we are speaking to do. Implimentations do exist, they must be above your target customer. As I said, this is not for the low end. To give customers pricing of 3, 4, 5, 10 PB's plus of storage on disk your customer will be doing all the laughing.

All the best..........

Well, that would have been useful information to know to better frame the context of the question. It was not meant as an attack, it was meant as sincere advice in the questionieare's best interest. But my apologizes to the questioner if it was taken otherwise.

Yeah and Edsels were great cars too.

You're right, I will never be convinced. I don't care who is/was using tape, it is still mechanical; still requires a controlled-speed drive mechanism; still has heads that touch the tape's surface and still is is subject to problems caused by dirt, dust and airborn pollution, unless you want to spend a boatload on clean room technology.

With the cost of disk storage continuously dropping (while tape is not) and with the advent of affordable high-density SSD storage just around the corner, the handwriting is on the wall for tape (and hard disks, too). I would guess that within 10 years at the outside (and probably a good deal less), mechanical storage will be an anachronism.

As Arthur C. Clarke said: "The ideal machine has no moving parts".

I wonder if hp has a Support Document for how to remove stuck hard drives like this:

"Stuck Tape Removal Instructions for Standalone Tape Drives"

We did a review of another digital tape drive solution, called Soleratec. Limitations and headaches in searching / retrieving is a problem. From what we have seen, size is one of two factors that are needed for digital tape to be attractive. The other is duration of storage. If you have 10PBs of storage but that's over a 2 week period, tape is a real pain in the ass because you are likely going to have lots of searches that need retrieving from offline storage. However, if your 10PBs is across 5 years and only 1% of searches are for video beyond the first month, digital tape's operational issues are minimized while getting the cost savings. Overall, digital tape looks to be a niche application even within large scale operations.

@John, I never eluded to that very short retention period. I think where it fits is in the large environment with longer term retention. Not necessarily 5 years. But in those applications where content is rarely retrieved as you said, less than 1% of the time after x days. It's green and to store content on disk which is rarely retrieved is a only good idea if money is no object. @Carl, you look at acquisitoin costs alone, OPEX counts too. Acquisition costs are greater than 50% savings today for tape vs disk. In PB plus environments less than 10 cents per GB. Not sure where enterprise disk is that cheap. I don't require another rebuttal, or insult, but I am not talking standalone tape and low end environments. Enterprise solutions, with proactive methods to avoid the issues you have seen with "old technology" and VCR tape. Media life cycle management that predicts errors way before they happen. The largest archives in the world are all doing this because they can't on disk due to the high acquisition costs and operational costs. As for HP, yea I'm sure they outline their disk rebuild process too. No technology is bullet proof.

Then why not just use a product like Veracity Coldstore in place of tape? You have a similar net benefit in terms of power and cooling savings and it does take a bit longer to spinup the disks containing the data of interest (but not as long as 2-3 minutes) without the negative aspects of tape.

Besides, I would bet most video surveillance only retrieves ~1% (or less) of the data actually writtten but when users do have to retrieve it, they often need it now, not in 2-3 minutes. Our retention times are 15 days, except for a few cameras at 60 days, and we randomly need to access anytime within that time frame. Storage quantity needs alone don't justify tape.

Where would tape be of benefit (and we will need close to a petabyte)?

Then you have systems that use DAS storage (which is many of the systems out there). If you have 20 servers, each recording and storing 100TB of data, would you have to buy 20 tape drives?

The benefit would be in the example above :) Thus the reason I brought it up, as I thought it had some validity. Veracity has a very nice product, but are their acquisition costs much lower than every disk manufacturer out there? It is no where near the cost benefit of digital tape, yet it has its place They are not the only game in town. Put in an archive server to share a few drives drives Carl, 10 cents a gig. Your issue will be can the computers deliver the content as fast as the tape drives can accept it. You're coming around, LOL. Every product has its fit. I have yet to have a customer say I can't wait 2-3 minutes to save greater than 50% acquisition costs. In the PB range, that is quite a bit of money. Customers just want the evidence from the investigation. It's interesting that the suggestions are how can we talk the customer into storing less. Lets throttle down the resolution, the frame rates, the retention. Why not give the customer what he wants? He did buy those 5 MP and 10 MP cameras for a reason.

Is it really only 2-3 minutes? That seems like a best case scenario if the tape is right next to the searcher. They do need to find the tape, get it, insert it, etc., no? What if they are in another building or city and need to call in and make a request? Then they need to switch to a different user interface and search for the specific clip they are looking for?

I am not saying it's never worth it but dismissing it as just 'waiting 2-3 minutes to save greater than 50% acquisition costs' seems to understate the operational issues involved.

Is grooming possible? ie store the full frame rate for one week and then drop it down to a couple frames for the remaining weeks. I know Ocularis ES (using the high end Milestone Recording Engine) can do this as well as Avigilon.

Everyone else brought up great points to look at. Personally if it was Arecont cameras I do not even think I would want to work on the project. We have had nothing but bad luck with those cameras and, unless they've changed, they "puke" out bandwidth with no control mechanism. If I absoultely had to bid this I would use Ocularis ES, Dell Recording Servers with either onboard storage or DAS for your "live" video (6-24 hours worth) and then a iSCSI SAN with it's own SAN network to archive the video long term. I would also get in writing from the VMS vendor that the configuration will work.

Jason, new Arecont firmware supports VBR with a cap to help control / limit bandwidth spikes.

Nice, good to know!

John this is late but undisclosed was not being attacked at all... There is a difference between an attack and a genuine concern. There should be an expectation of knowledge and expertise if someone is seriously considering implementing a solution of this size (with the customer in mind) this is NOT a learning experience project. Period. The naiveity of the subject question alone screams inexperience. Everyone must learn but this is a question that should NEVER be posed by an implementing party for a project of this size.

Joshua, I want these discussions to help members asking questions. I also want an environment that people feel welcome to ask more fundamental questions without being told that they are not good enough to work on such projects.

Secondly, you don't know enough about this integrator to make a judgment. You, and the other people who criticized this poster, should have taken a different approach if you were 'genuinely concerned', like asking questions: What's the largest system you have ever deployed? Have you ever integrated network based storage? What routing and switching expertise do you have, etc.? None of us know enough to make a determination about the overall skill of the integrator.

If you are really concerned, ask clarifying questions, don't jump to conclusions.


By my statement about the subject question my point is clear. This is not a question an integrator with experience would even pose. If someone wants to pose questions here everyone should have a chance to respond withOUT you taking a side. I can tell you most integrators who have implemented these type of solutions will agree - this is a question someone in the respective organization will KNOW the answer to based on the proposed solution.

The fact they are taking your class does not qualify them as there are far more up front (as you stated in your response above) and long term (actual support of system and infrastructure) concerns. Challenging or questioning someone who is diligent will not discourage them - it will simply strengthen their resolve to do it right... so don't baby your students - let them get the industry feedback they deserve to push them to be better.

Joshua, you are not challenging them, you are disparaging them, i.e., "The naiveity of the subject question alone screams inexperience."

If someone asks a technical question in the IPVM discussion group, I want it to be answered directly rather than 'challenging' their expertise. It's certainly worth explaining why something either cannot be done or requires a certain level of skill to do so, but your post was condescending. Again, I want less experienced members to feel welcome to ask basic questions and not feel personally criticized.