Retention of data should be determined by importance, one problem is we don’t know what will be important until it is needed.
Who determines what is important enough to keep?
If it is done by “AI” what are the rules?
If we assume 2/3 of the 15k police staff are “front line” i.e. those wearing body cams; that is roughly 10k cams recording 24/7. Each camera records @1-2GB/hr; this comes to 24-48TB/day of data; even if we mark 80% of that data for immediate deletion because nothing is happening (done by AI). We are talking 5-10TB/day that will need to be reviewed to determine if it is important enough to keep long term.
That is a lot of work; and a lot of resources to review the data. The storage quickly adds up; the benefits are huge, but the challenges of storage/security/privacy are equally enormous.
I’m not denying the enormity of the problem. My point was simply that something is better than nothing.
If we stored data for 30 days as standard, and if a complaint was made against an officer then that specific officer’s data was marked for longer retention, this could give us many benefits without needing to store much data long term.
Storing 10TB data a day for 30 days is 300TB continuous so a few grand a month of storage costs on backblaze. Similar on AWS if they use cold storage.
What if we didn’t reduce it by 80%. Let’s go with high numbers.
2GB/hour x 8 hours/day per officer. 250 working days a year roughly x 16GB is 4TB data per year per officer.
10,000 recorded officers x 4TB x $10TB/month storage = $400k per year for 12 months of storage, assuming deleting footage after 12 months.
$400k/year is couch cushion money for the government, they would easily save this avoiding one legal complaint by having body cameras. Did I miss a 0 or is this actually not as big of a deal as I thought?
Yes, the storage cost itself doesn’t seem to be the limiting factor. Presumably the body cam provider makes you use their storage which is jacked up in price but it seems you could find some middle ground (say, six months storage for most cops, some high risk roles get 12 months). And this provider storage would presumably come with the privacy and access requirements baked in.
I’m thinking this article is a softening the blow press release. They are announcing they are looking at it while already having decided they will do it, to get people used to the idea. Then later they will announce they are doing it in some future date, then finally turn them on.
Alternatively, they are signalling to providers they want cheaper rates for storage. Complaining publicly about the cost as a negotiation tactic.
Retention of data should be determined by importance, one problem is we don’t know what will be important until it is needed.
Who determines what is important enough to keep?
If it is done by “AI” what are the rules?
If we assume 2/3 of the 15k police staff are “front line” i.e. those wearing body cams; that is roughly 10k cams recording 24/7. Each camera records @1-2GB/hr; this comes to 24-48TB/day of data; even if we mark 80% of that data for immediate deletion because nothing is happening (done by AI). We are talking 5-10TB/day that will need to be reviewed to determine if it is important enough to keep long term.
That is a lot of work; and a lot of resources to review the data. The storage quickly adds up; the benefits are huge, but the challenges of storage/security/privacy are equally enormous.
Not sure why you were downvoted…
I’m not denying the enormity of the problem. My point was simply that something is better than nothing.
If we stored data for 30 days as standard, and if a complaint was made against an officer then that specific officer’s data was marked for longer retention, this could give us many benefits without needing to store much data long term.
Storing 10TB data a day for 30 days is 300TB continuous so a few grand a month of storage costs on backblaze. Similar on AWS if they use cold storage.
What if we didn’t reduce it by 80%. Let’s go with high numbers.
2GB/hour x 8 hours/day per officer. 250 working days a year roughly x 16GB is 4TB data per year per officer.
10,000 recorded officers x 4TB x $10TB/month storage = $400k per year for 12 months of storage, assuming deleting footage after 12 months.
$400k/year is couch cushion money for the government, they would easily save this avoiding one legal complaint by having body cameras. Did I miss a 0 or is this actually not as big of a deal as I thought?
I would assume that it would be something like 24TB/day [5k officers (2 shifts of 12 hrs) * 2GB/hr]; this would be for 365 days a year.
So yearly it is more like 8760TB for a years worth of data.
Cheapest I could find was Mega.nz at 16TB for $47.5/month which equates to ~$312k; this would be totally doable.
The processing requirements would be huge but a solvable issue also. Even if we double the data to 10k officers 24 hrs/day it is still under $650k/yr
Yes, the storage cost itself doesn’t seem to be the limiting factor. Presumably the body cam provider makes you use their storage which is jacked up in price but it seems you could find some middle ground (say, six months storage for most cops, some high risk roles get 12 months). And this provider storage would presumably come with the privacy and access requirements baked in.
I’m thinking this article is a softening the blow press release. They are announcing they are looking at it while already having decided they will do it, to get people used to the idea. Then later they will announce they are doing it in some future date, then finally turn them on.
Alternatively, they are signalling to providers they want cheaper rates for storage. Complaining publicly about the cost as a negotiation tactic.