Camera Edge Storage / Recording Tutorial 2015

JH
John Honovich
•Published Jun 29, 2015 04:00 AM

~70% of IP cameras allow storing video onboard, at the 'edge', typically through the use of SD cards.

The big question is whether and when it makes sense to use edge storage / recording inside the IP camera.

In this tutorial, we teach:

  • Edge compared to centralized and edge site storage
  • The 3 most common edge applications explained - server elimination, network load reduction and recording redundancy
  • Camera support and limitations
  • Third Party Compatability and ONVIF Profile G
  • Storage Reliability
  • Direct to NAS Alternative
  • Stealing Storage
  • The Business Case for Edge Storage
  • Future of Edge Storage

Compared ** *********** *** **** **** *******

****** *** ************ ******* ***** **** use ** *********** *******, ******* ** is * *** ****** ** ** NVR ** *** ********. ***** ** captured ** * ****** *** *********** to * '*******' ******** ***** ******** video ***** *** ******** ** *** same **, ****** ** *********.

****** *************, **** **** *****, ********* have * ******** (** ********* ********* on ****** ** *******) ** **** site. **** *** ** ********** * form ** '**** *******' *** ** is *** ****** ***** **** ******* but ********* ******** ** *** ****.

** ********, **** **** *******, *** video ** ********, ******* *** ****** right ****** *** ****** ******. ***** does *** **** ** ** **** out ** *** ****** ****** ******* wants ** **** **** ***** ** review ******** *****. *******, **** ** cameras ******* **** *******; **** ** not ******** ** ****** ** ** SDI ******* **** **** ** ******** their ***** ** *****-***** *********. ** cameras **** ****** ** ***** *** local **** ******* *** *********** '****** channel ****" ** **** ******* **** video ******* *** ********* ** ***.

***** ***** *** ** ******** ** many ******* *******, *** **** ******* are** *****:

  • ** **** ***** *** ********** ***** into ******* ******** ******* ** ** plugged **.
  • ** *** ****, * *** ************* (e.g. *******) ***** **** ****** **** the ******* *** *** *********** ********* size ************ **** **** *********.
  • ** *** ****, * *** ************* allowed ******** **** ***** ** ** connected ******** ** *** ******.

Common ************

***** ********* ************ *** **** ****** for **** *******:

  • *** ****** ***********- ******* ** ******* ** * server ** ********* ** ***, **** record ****** *** ****** ***** **** storage *** ******* ** ***** ******* directly ***** * ******. **** *** the **** ****** ** ***** **** a ***** ****** ** ******* ** VMS ******* ********* **************** **** *** complexity ****.
  • ******* **** *********- *** *** ********* ********, **** wireless, ********* ****** *** ****** ********* reduces *** ******* ** *** *******. Instead ** ********* ************ ** ******** there ** ******, *** ***** **** is *********** **** ******* ***** ** watch **** ** ** ** *************.
  • ********* **********- **** ******* *** ** **** as * ****-** ** ********* *********, ensuring **** ** ***** ** **** even ** *** ******* ********** **** down *********** ** *** *** / NVR ** ******* ************. **** *** connection ** *** ******** ** *************, the ******** *** ******* ***** ****** during *** ******** ** ** **** from *** ****** ** ** (****: must ****** *** ******* *** **** feature *** *********** **** ******** *******). Redundancy *** **** ******* ** **** often **** ** ***** *****, ******* critical ************.

Camera *******

************, ** **** ******* ** *********** for ** *******, **** ** ***-****, low-end ****** ** *********** ** ********** cost.**% ** ** ************** ******* ******* ** *** **** Camera ****** , ****** *** *** SD *****.

***** *** ***** ********* *********** ********** encountered:

  • ******* ****
  • ******* ***********
  • ******** / ********* *************

Storage **** ***********

** ****, *** *** ******* **** per **** **** ** ** ******* ranges **** **** ** *****.

**** ** *******, ********** ***** ****, only ******* **** *****, ******** ******* to ****. ***** ******* *** ******* SDXC, ***** ******* **** **********. *******, regardless ** *** ** **** ******* supported, **** **** *** ************ ******** the ******* ******* **** **** *******.

** ********, **** ** ******** ****** common ***, *** *******, ** **** Axis *** '************' ** **** ********* ******* ***** ** *****.

**** *** ****** ** ********** ** a *** ** ******* ********, ***** for *****'* ** ***** ** *** a *** ** *******. **** *** camera, *** **** ** **** ****** depends ** **********, ***** ****, ***********, motion ** ********** *********, ******** ******, etc. **** ****, ******* * ***** of ********* **** ** **** **** without *********** ** *** ** ***** key ********** *****.

Storage ***********

** *** **** *** *****, ******** over **** ******* *********** **** ****** very ***********. *** *******, ***:**** **** ******* *** ****** ********* Is *************** ****** **** ********* ********, ****** Solution. **** **** *** ******* ***, at ***** ********* ******* **** *** a ***** ******* *** ******** **** storage *******. *******, **** ******* ** issues **** ***** ******'* **** ******* also **** ********.

*** ***** ****** ** *** *** of ***** '*******' ** *****, ** SD ***** ***** **** ****** ***** giveaways ** ********** *********. **** ******** people ********* ***** ** ** *****, though ** ******* **** ** ************ as '*****' **** ******** **********, **** part ** **** ** ******** *** reliable *******. ********, *******, *** *** taken * *** ** ********* *** its ***** ** ************,******** * '************ *********' *******, ****** **** ** **** *** and **** **** **** *.* ***** effective ********.

****** ****: ******* *********** ** * concern *** ****** *** ****** **** storage, ***** **** **** ******** ****** supplier ** **** **** ********* *** are ******* ** ******* / ********.

Third ***** *************

*** ******* ******* ***** **, *************, the **** ** * ** * years ***. ***** ***** *** ******* for **** ******* ******* ****. **** if *** *** ****** ***** ****** a ******, ** **** *** **** you *** ****** ******** **. **** IP ******* ******* ** ***** *** only ***** *** *********** ******* ** individual ***** *****. **** **** ******* VMS *********** *** **** *** ***** own ***, *** *** ***** *******.

** *** *********** ****, ***** ********* expect ** ******* ****** ** ************'* cameras ** * *** *** ****** them ***. *******, **** **** *******, most ***** ******* ** ******* *** the *** **** ** ******* *** party ****** *************, ********* **** ****** Axis. **** **** **** *** ****, Genetec ****** *** *** ****** ***** party **** ******* ******* *** **** is ***** * ******** ** *** cameras.

*** / ***: ****** **** **** choice ** ********* *** ***** ********* with **** ****** ** **** *******.

ONVIF ******* *

* ***** ***** *******, *, **** to ******** ******* / ********* ** recorded *****, ** *******. **** *** the ********* ** **** ***** ***** systems ********** ** ***** / ******** as **** ** ***** / ******** integrating **** ****** **** *******. ** the **** ****, ******** ****** ******* for ***** ******* * ** *******, with *** ****** ** ** **** 2015. *******, ** ***** *** ********** supporting ***** ******* * ** **** time, ************* ************ ******** ************** ** ***** * ******* ** it*** ***** **** *******.

******* * ** ****** ******** *** not ** * **** **** ****. The *********** **** ** ***** **** today ** **** ** *** *** it **** **** ********* ** **** reduce *** ***** ***** ************* *******.

Edge ******* *********** - ****** ** ***

*** ***********, ****** ** ***, ****** be ********** *** ***** *** **** to ********* *** ******* *** ****** less ********* ***/** ******* ******* ** storage. *******, **** **** **** **** if ***'* ****** *** ******** ****** to *** *********, ***** ** ********. See ********* ** *************** *** *** **** *** **** of ****.

Stealing *******

***** ******* *** ****** *********** ****** and ******* ** **********, ***** **** storage ***** ******* *** ******** ***** wherever *** ******* ***. *******, **** makes ** ****** *** ** ********* to ******* ***** ********. *** **** need ** ** ** ****** *** back ** *** ****** *** **** out *** ****. *** **** ** a ********* **** **** ** *** whether ***** **** ****** ** **** vary ***** ** *** ******** ******** of *** ************. ** ****** **** high ******** ********** ** **** **** as ************ ***** *** **** ** the ****** ** ** **** ********* about ****.

The ******** **** *** **** *******

*** **** ******** **** *** **** storage ***** ** ******, **** ******* systems **** ***** **********. ** ***** systems, ***** ***** ********* ******** ** corporate ** ********, *** *********** ***** of * ~$*** ** **** *** camera ** *** *********** *** *** likelihood **** *** *** ***** **** supports **** ********* ** ******, ****** not *******.

* *** ***** ***, ***** **** storage ** ********* *********** ***** / recorders *** ********** *** *** **** has ****** ****** *****. *** ***** who ****** ****** *** ****** ****, HD ****** *** ******** * ******* and ***** **** *********** ** ******* this. **** **** ********* **** **** Camera *********, ***** *** ****** **** of **** ******* *** *** ~$*** SD **** **** *** ****** ****** come ***** ** *****, ********** ***** HD ****** **** *** ** ***********.

*** *****, ******, ***** **** ******* replacing ***** *** ****, ** ** only * *** ******* (*, * or *) ** ****** *** ****. In **** *********, *** **** *** complexity ** ****** * ******** ********* a *** ** *****.

Future ** **** ******* / *********

**** ******* ** ********** * ***** in **** (******* **** ******* ***** **********).

*** *** ******** *** (*) ***** party *************, (*) *********** *** (*) size. ****** ***** *** **** **** progress ** **** ***** **** *** past *** *****, ** ** *** fast. ******, *** ******* ***** ******* looks ** ** ** * ******* pace. ** ****, ******* **** ******* when ***** **** *******.

****: **** ** * ************* ******* version ** ******* **** ******* / ********* ********.

Comments (16)
DW
Dale White
•Jun 29, 2015

John,

Thanks for this. Maybe the manufacturers on here will move toward improving the offerings in this area. I have advocated for edge storage for a long time. Many other technologies moved intelligence and data to the edge years ago e.g access control. At one time, all access control decisions were made by a central processor which limited response time etc., but that was 30 years ago.

To me, I guess it's just logical to keep your data locally and only access it when needed. I wonder if use has improved since your 2012 article?

We spend a lot of money shipping huge amounts of video data around to massive storage arrays, clouds, etc. Our studys have shown that <2% of video is ever actually used. So at a site with 24TB of storage you're just hogging the network infrastructure for no reason.

That's like having a 15 ton dump truck deliver a shovel full of dirt. We have a ways to go in this area but the time could be short with the right company and resources.

Given the average system size, reliable, supervised edge storage would simplify a lot overly complex installations. Not sure why it hasn't been a priority before now.

Can't wait!

JH
John Honovich
•Jun 29, 2015
IPVM
In reply to Dale White

Dale, nice feedback!

"I wonder if use has improved since your 2012 article?"

We have 2014 edge storage statistics here - not much adoption and a lot of barriers cited.

"We spend a lot of money shipping huge amounts of video data around to massive storage arrays, clouds, etc."

Maybe I do not understand how you are using these terms. Indeed, there might be 3 ways here:

  • 'Pure' or 'Strict' edge - which is storage inside the edge device, i.e., camera
  • 'Site' or 'Store' edge - where video is moved into centralized recorders at the site or store but distributed relative to the company / organization (i.e., 100 sites has 100 recorders, 1 at each site).
  • 'Pure' or 'Strict' centralized - where no video is stored at the edge at all, and all video is moved across the WAN to a single data center (either internal to an organization or in the 'cloud')

The 'site' or 'store' edge has been the most common solution, because of the tradeoffs of going to either 'pure' position. Is that what you are alluding to or?

DW
Dale White
•Jun 30, 2015
In reply to John Honovich

Today we store as in your "Site" edge example so the site network has all of the camera traffic all of the time with a large storage platform. Since for most purposes, a site is a system, I guess I don't consider that edge.

There is a recurring discussion to consider shipping video from all sites across a WAN to a central storage. As in your "strict" centralized. Yikes!

When I'm thinking edge, I think of "Strict" edge as in a modern badge reader that stores the data locally and can function independently of the host. Storage in the camera. Access only the video needed. If we can get to a reliable 128GB+ card that is supervised by the VMS, I dont see any reason why this can't be a viable architecture for small 50 cameras or less systems.

JH
John Honovich
•Jun 30, 2015
IPVM
In reply to Dale White

"If we can get to a reliable 128GB+ card that is supervised by the VMS, I dont see any reason why this can't be a viable architecture for small 50 cameras or less systems."

I think you can get there in the next few years. But what is the big benefit? Eliminating the traditional recorder on-site?

I wonder about:

  • Maintenance logistics - how much more cost, complexity, headache is dealing with dozens of SD cards vs a few hard drives?
  • Cameras needing more storage - some cameras because they cover a more active area, will require more storage for the same time requirements which could be an issue (i.e., all cameras have 128GB cards, some cameras record for 50 days, others only record for 15, etc.)
  • Accessibility issues - will retrieving recordings from many different devices cause delays or issues in playback? What happens when a camera is inaccessible, which is typically more common than a recorder.

I am not against it, I am just not sure if it will be a big win for larger organizations to go 'pure' edge.

DW
Dale White
•Jun 30, 2015
In reply to John Honovich

I think you're right about the larger organizations. Even so, large organizations have many intances that are small, "one off", or temorary sites where a big head end just isn't justified. Places that have no security or monitoring presence at the site (remote access only). Also, small sites that are "tethered" to a larger facility.

On an enterprise scale pure edge may become unmanageable, for the reasons you mention. For a small or medium school district, or convenience store chain, maybe not.

I think the biggest obstacle has been the manufactuers haven't taken it seriously. If I'm able to store 250 or 500GB of data on a camera and I'm only ever going to view <2% of it, why would I want to continuously move all of it over a network somewhere else to store it?

JH
John Honovich
•Jun 30, 2015
IPVM
In reply to Dale White

I agree about small sites.

"If I'm able to store 250 or 500GB of data on a camera and I'm only ever going to view <2% of it, why would I want to continuously move all of it over a network somewhere else to store it?"

To play devil's advocate:

(1) You need a LAN connection anyway to the camera and whether you are sending 1MB once a day or 8Mb/s, on a LAN the incremental cost is minimal (i.e., sending from the camera to a recorder on site).

(2) The cost of storage in a box (recorder, NAS, etc.) can be less expensive than the cost of storage inside a camera (micro SD card).

U
Undisclosed #1
•Jun 29, 2015

Some good points, a few things I'd mention as well:

The two big limitations of edge-storage are overall card size (in GB) and difficulty in accessing the video (both of these are mentioned, I'm just confirming/rehashing). Edge storage is great for cases where you only have a small number of cameras and will need to access/retrieve recorded video very rarely.

I disagree with this statement though: This is not possible in analog or HD SDI cameras that need to transmit their video to stand-alone recorders.

Edge storage is really just storing the video in a way that keeps the traffic off the network. Cameras directly connected to an NVR/DVR would really be edge storage by most practical definitions.

Something else not mentioned in the article is the use of edge storage for network load management. I see/hear people ask about this a lot, and it doesn't work like that. Generally speaking, if you don't have the bandwidth to record video live, you won't be able to "buffer" it in the camera and move/trickle it to some centralized recorder later.

In terms of SD cards, I'd strongly suggest budgeting for a 3 year replacement cycle. You can use the data sheets from most of the popular cards and some basic math to calculate an expected life span, but in my experience there is more to it than just the total number of write cycles.

The good news is that bad SD cards will tend to show their flaws pretty early, and you can catch some of this before installing them in the field. Essentially:

1) Determine how much data you will use per day (within a few GB or so, it doesn't have to be perfect). Let's assume you're storing 30GB/day.

2) Multiple your daily storage by 90(days) for a total, 2700GB in this case.

3) Take your SD card size and divide the number above by your card size (eg: 2700GB/128GB card = 21.09)

4) Create a set of 3 random files, each that is a little bit smaller than your card size (you're going to copy this file to the card and you want to make sure you don't get an error from the file being too big for the card after your OS puts some crap on the card first).

5) Mount the SD card, copy thr first file, when that completes delete it and copy the 2nd, then the third. Do this process however many times simulates 90 days (eg: 21 times in this case).

6) Use a utility like this one to check the card for errors. If there are a high number of errors discard or return the card.

Or, to state the above another way, I've seen a lot of cards fail in the first 60-90 days, if they last that long they'll probably last a couple of years. You can simulate a bunch of write-cycles with some script-fu on linux or Windows and catch faulty cards before they're 100 miles away and 20 feet in the air.

JH
John Honovich
•Jun 30, 2015
IPVM
In reply to Undisclosed #1

"I disagree with this statement though: This is not possible in analog or HD SDI cameras that need to transmit their video to stand-alone recorders."

You disagree because, as you acknowledge, you define the term 'edge storage' differently.

"Cameras directly connected to an NVR/DVR would really be edge storage by most practical definitions."

This article, as the title notes, is about 'camera edge storage' in particular. NVR/DVR edge storage is perfectly valid and far more commonly used than camera edge storage, simply a different thing.

"Something else not mentioned in the article is the use of edge storage for network load management."

That is covered in point 2 of common applications.

Good advice on the SD card tests. Would be interesting / potentially useful for surveillance manufacturers or distributors do this as a service.

U
Undisclosed #2
•Jun 30, 2015
IPVMU Certified
In reply to John Honovich

Good advice on the SD card tests. Would be interesting / potentially useful for surveillance manufacturers or distributors do this as a service.

Do you think it's worth the 90 days additional wear you put on the card though? Maybe just a couple out of a batch would be enough, no?

JH
John Honovich
•Jun 30, 2015
IPVM
In reply to Undisclosed #2

Not my speciality. Or they could just go with the SanDisk surveillance cards and hope / base their performance on that.

It's a tough call. I still think camera manufacturers should make it as easy as possible to get the SD cards from them, to reduce any dumb problems like a guy grabbing whatever card is at the bottom of their office draw, etc.

U
Undisclosed #2
•Jun 29, 2015
IPVMU Certified

Or, to state the above another way, I've seen a lot of cards fail in the first 60-90 days, if they last that long they'll probably last a couple of years. You can simulate a bunch of write-cycles with some script-fu on linux or Windows and catch faulty cards before they're 100 miles away and 20 feet in the air.

If I'm understanding you correctly, this seems odd. I don't doubt that there is a good deal of variance in reliability of cards even of the same type and capacity, but it is well known that these cards lifespan is limited by the total number of cell rewrites, at least for the working ones.

Needlessly simulating 90 days of writes would be expected to reduce the lifespan of the card by 90 days on the ones passing the test. Do alot of them fail in just 2 or 3 days? Maybe that would be enough to eliminate the worst cards.

I disagree with this statement though: This is not possible in analog or HD SDI cameras that need to transmit their video to stand-alone recorders.

Why disagree? I suppose that SDI cameras could record uncompressed video to the SD card, but I would imagine that would fill up in a matter of minutes. As for any analog camera, I don't see how its possible at all.

U
Undisclosed #1
•Jun 30, 2015

Needlessly simulating 90 days of writes would be expected to reduce the lifespan of the card by 90 days on the ones passing the test. Do alot of them fail in just 2 or 3 days? Maybe that would be enough to eliminate the worst cards.

It would *sort of* reduce the lifespan, but not as directly as you think.

The cards have a spec that lists the number of overwrite/bit flip cycles before major degradation, and loss/corruption of data, occurs. The majority of the cards will fit to this spec (actually, exceed it) and everything is OK, but some will not.

I'm not saying this is an exact science, it's just some recommendations based on observations. 2 or 3 days worth of testing would only equate to 1 or 2 complete flips of the card data, almost no name-brand card should fail that quickly.

This is really no different than the burn-in tests typically done with components. Due to the nature of the SD cards being installed in the cameras it's generally not easy to just walk up and replace the card. Also, because of the way they fail, and the fact that many cameras still don't do much to check card integrity and report errors (last time I looked, over a year ago), you can end up in a case where your card is either completely dead, or at a very reduced capacity. If you are not regularly accessing stored video to verify the integrity of the card you will end up in a case where when you need the video it's not going to be there.

If the card can survive a simulated 90 day stress test there is a higher chance that it will go on to be reliable. I also feel that if you're using SD cards exclusively you need to setup a maintenance window that gives you a good margin of error. If a 90-day reduction in lifespan (and again, it's not direct) is going to be a major factor, then you should be using larger cards or not relying on edge storage for your solution.

(1)
U
Undisclosed #2
•Jun 30, 2015
IPVMU Certified
In reply to Undisclosed #1

This is really no different than the burn-in tests typically done with components.

In theory it's similar, though most components I'm aware of burn 24/48 hours, not for months. Most of the reason for that though is just because of the pipeline delay cost of having them there longer.

I'm not saying it doesn't make any sense, it's just that I feel like when I put an SD card into a continuously writing cycle, that the clock is ticking. But I can see where you are coming from.

Do you think it might be helpful to 'burn' a number of cards until they fail? That way you could graph where the biggest valley of reliability lies and adjust accordingly.

AG
Alexander Gutierrez
•Jul 09, 2015

My experience with Sunell OEMs when using Edge recording (on Micro SD Slot): For a strange reason the Clock resets, so the recording is affected, either the camera does not record or the recorded file has a false Time/Date info. This need to be checked when testing these kind of solutions. Sunell provides a NTP tool but this is useless when you want to apply to real life situations because you are only installing a camera with no local VMS / NVR.

JJ
Joe Jones
•May 27, 2022

John,

I was wondering if there has been any newer research done into edge recording since this article. I know that the storage capabilities have evolved quite a bit and was wondering about 3rd party support and other features.

JH
John Honovich
•May 27, 2022
IPVM
In reply to Joe Jones

We don't have anything newer now but we are planning to add a new direct to cloud camera storage guide to cover the biggest development in this area.

(1)