[gruez]

>Also, several minutes of buffer fits nicely in RAM. 5 hours would require a lot more RAM (making devices more expensive and thus harder to get deployed) or would require it to be written to flash disk which introduces new technical and legal issues.

If you're concerned about writing potentially sensitive video to non-volatile storage, that can be solved via technical means. Just encrypt the video with an ephemeral key kept in memory, then store the encrypted video on non-volatile storage. You get best of both worlds.

>I've done that math and the numbers are staggering.

Mind showing your calculations? Keep in mind you don't have to store the videos in perpetuity. Limiting it to 180 days (with option to extend if the incident is being disputed) would suffice for most cases.

[davidf560]

> If you're concerned about writing potentially sensitive video to non-volatile storage, that can be solved via technical means. Just encrypt the video with an ephemeral key kept in memory, then store the encrypted video on non-volatile storage. You get best of both worlds.

Once you've recorded that video it's subject to FOIA requests and rules of evidence and all kinds of things like that. If your recording only includes the past minute or 2, you can be reasonably sure it doesn't include something it shouldn't (such as a victim who asked not to be recorded from an hour ago). If hitting that button commits a 5 hour buffer, then you're going to have all kinds of stuff now "in the public record" that maybe shouldn't be there. But also you now have 5 hours of more or less useless video that you have to store and account for (which costs $).

> Mind showing your calculations? Keep in mind you don't have to store the videos in perpetuity. Limiting it to 180 days (with option to extend if the incident is being disputed) would suffice for most cases.

It's actually less about the storage and more about the bandwidth to transfer that data to the storage location. Though the cost of the storage itself can still be an issue even if you're only retaining it for 90-180 days.

There's a whole bunch of factors of course, but pick a reasonable bit rate (5-10 Mbps perhaps) for your recording and multiply it by the number of hours recorded per officer and the number of officers in an agency. Chicago PD has 12,000 police officers. I don't know how many of them work on a given day but let's say 10,000. Say each one records 5 hours per day based on the 5 hour buffer proposed here.

10,000 * 5 hours = 50,000 hours of footage per day

50,000 hours * 5 Mbps = 112.5 terabytes per day

112.5 terabytes uploaded in 24 hours = 10.4 Gbps

(you need to be able to upload a full day's recording within 24 hours or you'll never catch up)

Doing this quickly so hopefully my math isn't off but it's in line with what I remember. 10Gbps being uploaded 24/7 to the cloud. 20 petabytes stored at any moment if you're holding everything for 180 days.

Sure, you can attack some of these numbers (not every officer will have a recordable interaction every day etc.) but I hope it gets the point across.

[gruez]

>10Gbps being uploaded 24/7 to the cloud

that might seem a lot if there's only one server, but it's really not that much considering that it's going to be uploaded from multiple precincts. Wikipedia says there are 21 "districts", so that works out to a 500 Mb/s link per building, which seems pretty modest for a institution.

>20 petabytes stored at any moment if you're holding everything for 180 days.

If it's being stored using aws glacier deep archive, that's only $20k/month. That's probably a drop in the bucket compared to the rest of the costs associated with running such a service.

[davidf560]

> that works out to a 500 Mb/s link per building, which seems pretty modest for a institution

We're talking 500 Mb/s of upload 24/7 though, which can't be used for anything else (i.e. you need capacity beyond that for any other internet traffic). I doubt a regular old 1Gbps business class connection will cut it there (I'm sure they'd shut you off after a few days of uploading like that).

Despite another post that cited a large budget for NYPD, most agencies are cash-strapped. Telling them they need a $thousands/month internet connection at every precinct isn't going to go over well.

I've been to several PDs (think suburban city - probably like the majority of agencies) where their entire "internet infrastructure" consisted of a Comcast cable modem screwed to a sheet of plywood on a wall. Like I said elsewhere, most PDs are the opposite of high-tech - compared to most companies, their employees don't spend nearly as much time on computers or the internet.