FreeNAS as DVR for IP cameras?

[Black Ninja]

So, the key point they're making is that a DVR that may claim to support 1080p, but it won't be able to support that at 30 FPS for the full number of channels. The solution is to pay more for a DVR that can support greater bandwidth, reduce the number of channels, or reduce the resolution.
There's another option. Reduce the FPS. You're not producing a feature film here and there's no need to record security feeds at 30 FPS. Drop your camera FPS to something around 7 and you'll easily fit all the channels you need. Even dropping the FPS to 15 would suffice instead of dropping to 720p.

Probably you are talking about analog cameras, not IP cameras, otherwise non of that will apply.

 

[Black Ninja]

@Rickinfl I notice from the screenshot that you are using separate dataset for each camera, I wonder why is that? Instead of just one shared dataset for all cameras?

Also what advantage you see by using a NFS instead of plain FTP for access to your FreeNAS?

 

[tvsjr]

You are right, it could be a problem if the memory is spread in all the banks , but with few sticks it shouldn't be.

I know "L" is running on slower clock , but I don't know if that's simply and the only the difference that gives it such a low power consumption compared to the non "L". I can't speak from experience (comparing both on idle)
but that could be the case. Very good point.

P.S. You do have a nice NAS in you signature, I would assume you don't have much use of second CPU , but perhaps you got a great deal on the system as a whole , that's the reason is dual CPU , right ?

No, I built it that way. It loafs along most of the time but if I hit it with serious iSCSI traffic (like during a vMotion) I'll see load averages of 8-10.

And - there is no kill like overkill!

 

[Black Ninja]

No, I built it that way. It loafs along most of the time but if I hit it with serious iSCSI traffic (like during a vMotion) I'll see load averages of 8-10.

And - there is no kill like overkill!

I see. So 8-10 % total across the both CPUs. So if this was single E5-2670 will probably still not go above 20% ?!

I think I FILL the OVERKILL here :)

I do like overkill, by the way. But wonder if single E5-1650v2 won't make it a faster machine?! I wonder what is the power consumption on this machine?

 

[tvsjr]

No, Linux load averages don't work that way. A fully loaded CPU will display a load average equal to its core count (so for my box, 16 would be 100% loaded). We'll ignore hyperthreading for this discussion...

 

[Black Ninja]

No, Linux load averages don't work that way. A fully loaded CPU will display a load average equal to its core count (so for my box, 16 would be 100% loaded). We'll ignore hyperthreading for this discussion...

You mean FreeBSD.:)
Anyway I know what you are saying, I thought you were looking at CPU GUI graph. I am not sure is this case does it take in consideration the clock state the cpu is. I mean in windows if you allow CPU to down clock with the C states, and stays at 1800Ghz and some process jumps cpu load can show is 20-30 % but that's out of 1800Ghz , not at full speed which is let say 3Ghz, so it's hard to tell. Well my workstation stays at 3.9 Ghz flat all the time, so I can tell when it jumps 10% that's 10% out of my max speed.

So using your data from let's say "top" for the cpu load as you suggest. So having average of 8, should mean that your system in that moment is 50% loaded all across the CPUs , is that right? (Which is 2 CPUs with 8 cores each)

 

[tvsjr]

I meant *nix, but yeah, you got the idea. And yes, a load average of 8.00 would be 50% CPU utilization. Again, ignoring hyperthreading (you could conceivably see a load average of 32 on a 16-core box, but that assumes a workload that is 100% friendly to hyperthreading).

Do I need the amount of CPU I have? Not hardly. But I've got it and it doesn't consume massive amounts of power just sitting there idling.

 

[Black Ninja]

But I've got it and it doesn't consume massive amounts of power just sitting there idling.

I am just curios how much power, cause "not much" is a question of prospective. For example the FreeNAS in my signature. Let me log in right now and I'll tell you.... so mine is 75+89= total of 164 Watts that's on pretty much idle. I find that very good.

 

[Black Ninja]

I was just curious because I was actually interested in getting Supermicro CSE-847E16 chassis. :)

 

[tvsjr]

My system idles around 325 watts. But, the majority of that power is consumed by the 18 7200rpm drives I'm running.

 

[Black Ninja]

My system idles around 325 watts. But, the majority of that power is consumed by the 18 7200rpm drives I'm running.

Thanks. I count every watt, because few servers and add up and get you over 1000W and I am trying to stay in 1000W range.

 

[cobrakiller58]

Probably you are talking about analog cameras, not IP cameras, otherwise non of that will apply.

Are you implying that you cannot reduce the FPS on an IP camera to record more channels at higher resolution?

 

[tvsjr]

Thanks. I count every watt, cause few servers and add up and get you over 1000W and I an trying to stay in 1000W range.

My whole rack runs around 1,100W and will jump to 1,600-1,700W if I really hammer it. That includes:
FreeNAS server (in my sig)
3x 16-core vSphere nodes
2x 4-core domain controllers
1x 4-core pfSense box
2x gigabit switches (one PoE, powering 11 IP cameras and 3 wifi APs)
All the various little stuff... cable modem, LTE modem, distribution amp for antenna TV, HDHomeRun boxes for antenna and cable, etc.

 

[fracai]

Probably you are talking about analog cameras, not IP cameras, otherwise non of that will apply.

are you implying that you cannot reduce the FPS on an IP camera to record more channels at higher resolution?

Right, this has me confused. I definitely have my IP cameras configured to run at 7 FPS. That's a configuration of the camera. I suppose it's possible that the camera is still putting out 30 FPS worth of data (I will admit that I had not considered that possibility and have not checked for a change in throughput), but that sounds like it would be a phenomenally bad design.
Even if it turns out to not change the required bandwidth, reducing the FPS does change the CPU time required to analyze, which is a consideration in my case.
I'll run an experiment to see how the throughput changes with my cameras.

 

[Black Ninja]

are you implying that you cannot reduce the FPS on an IP camera to record more channels at higher resolution?

I am implying that is irrelevant. Your DVR (FreeNAS serving as DVR) accept steams of digital data (FTP/NFS) it doesn't care about your FPS, resolution, color, etc. it only cares about how many clients send data and how fast , and is writing in to a files.

Perhaps you are confused somewhere along the conversation, or perhaps we are talking about different things.

 

[Black Ninja]

My whole rack runs around 1,100W and will jump to 1,600-1,700W if I really hammer it. That includes:
FreeNAS server (in my sig)
3x 16-core vSphere nodes
2x 4-core domain controllers
1x 4-core pfSense box
2x gigabit switches (one PoE, powering 11 IP cameras and 3 wifi APs)
All the various little stuff... cable modem, LTE modem, distribution amp for antenna TV, HDHomeRun boxes for antenna and cable, etc.

That's exactly what I mean by things (watts) add up quickly. :)

I knew your system was not alone in the rack, same as mine. From the power outlet to the rack I have 3 UPS's and each is taking around 15-20 Watts before anything was connected to it.

My pfSense box 1x8 cores Atom C2000 16Gb Ram, 2x120 SSD is burning only 34 Watts. Which is absolutely great, considering that it resides in 1U chassis with dual power supplies.

It's nice when you are able to make machine more power efficient so you can run more machines :)

Just by switching my NAS (signature) old dual 800W PSU with new dual 500W platinum , machine became much quieter and at lest 40W more efficient, so now I can add this 1U DVR dedicated FreeNAS server (39W idle with drives) and keep total power consumption the same.:)



P.S. Your power consumption look to low from what you are describing, are you sure you measure at the tap?:)
Because this section alone :
"FreeNAS server (in my sig)
3x 16-core vSphere nodes
2x 4-core domain controllers"
feels like 1Kw just there. But it is very good if idle around 1100W for what you describe.

 

[tvsjr]

I have a Cyberpower metered/switched-by-outlet 16-port PDU that runs everything... so I have accuracy down to 10mA of the current consumption of every port. Plus a Cyberpower 3KVa online dual-conversion UPS running everything. So I know *exactly* what everything uses, and can chart it all over time too :D

 

[Black Ninja]

This sound cool and interesting. So you run everything on a single circuit on single UPS right?

P.S. I had a intention of getting dual-conversion UPSs , but then I got convinced that it won't increase my reliability in any way but will decreased my power efficiency significant. But of course if you live in a area where huge fluctuation in power exist, you have to use dual-conversion.

 

[tvsjr]

Yes. When we remodeled this home, I had two dedicated 120VAC/30A circuits run to my server closet. One is dedicated to the UPS, one powers the battery charger for the extended battery pack. The UPS I have can run in two modes... in ECO mode, it's sorta semi-online - it can be >95% efficient in this mode. I don't run it in that mode... while we actually have very good power here (it's all underground, the transformer is actually in the corner of my front yard, etc.) I still like the extra protection.

 

[Black Ninja]

I had two dedicated 120VAC/30A circuits run to my server closet. One is dedicated to the UPS, one powers the battery charger for the extended battery pack.

So 2 circuits, but one UPS? Not sure what "battery charger for the extended battery pack" would be as stand alone devices, but basically no redundancy of UPS failure, right? So if on the first circuit the UPS goes down, everything goes down, right?