We need to place 16 IP cameras (4MP) to monitor a warehouse, where all these cameras will be handled by a CCTV. We have 2 options:

  1. Connect these cameras through the same cables under the same switch where our PCs are connected..
  2. Place new wires and a new switch to and connect all the cameras and IPs camera there. The 2 switches are PoE (cameras will be powered on from the switch) and will be under the same subnet.

enter image description here

Is option 2 a better plan to follow, in order to reduce bandwidth in our network where PCs are connected? Or the traffic that cameras are sending to the CCTV device will not increase significantly the bandwidth?

*Switches are gigabits (are managed,the wires CAT6 and IP cameras have 4MP resolution . The CCTV has a single port in order to be connected to the network

Cameras have Video Bit Rate : 32 Kbps to 8 Mbps.

In the switch there are 5 office PC, which have access to a local NAS disk for shared files (excel, word, pdf most common files) and 3 Access Point for 20 wireless devices (mobiles, tablets).

Specs of the switch:

  • Total Non-Blocking Throughput: 70 Gbps
  • Switching Capacity: 140 Gbps
  • Forwarding Rate: 104.16 Mpps
  • 1
    In picture B, is the router routing between the two separate networks ? Or is the router using two switchports in the same network to connect two physical switches in the same logical network ?
    – Criggie
    Mar 15, 2020 at 3:11
  • 1
    @Criggie I edited the question
    – yaylitzis
    Mar 16, 2020 at 6:23
  • 1
    No, the cameras are powered only from the switch. I will not use any other power wire. The switch has PoE ports.
    – yaylitzis
    Mar 16, 2020 at 6:35

4 Answers 4


The frame rate of the cameras will determine the data rate. If the data rate is between 32 Kbps and 16 Mbps, the bandwidth for 16 cameras is between 256 Kbps and 128 Mbps.

We can't tell you if that is acceptable or not -- it depends on whatever else you have on that switch.

At the low end of the range, that bandwidth used isn't significant. At the high end it also probably isn't, but might be.

You will have to trade off video quality for bandwidth, but only you can determine whether the trade off is acceptable.

  • Again, we can't determine whether this is acceptable, and opinions are off topic. But based on the specs, you can run all your ports at line rate.
    – Ron Trunk
    Mar 14, 2020 at 14:10

16 Cameras at 4MP resolution, running at 15 fps gives the following:

MJPEG (medium): 941 Mbps
H.264 (base): 63.7 Mbps
H.265 HEVC (medium: 25.6 Mbps

Also remember that even if you're doing motion-only recording, the cameras will constantly be sending a stream to the recording NVR. Also, any live monitors will duplicate the bandwidth from the NVR to the client if they're running over the network.

As a side note: Consider running the cameras on H264/H265 if at all possible, as you can see this will save you quite a bit of bandwidth compared to MJPEG. The network is however seldom the bottleneck with IP cameras, it's usually the recording device CPU, NICs, or disc system that gives out first.

Option B also has the disadvantage of directing traffic from viewing PCs through the router, which might quickly become saturated as routers are quite bad at passing traffic compared to switches, especially the lower end ones.

  • 1
    That last paragraph assumes the router is routing two different IP networks. It could be a generic "router" with a ~4 wired LAN ports, so a third switch rather than routing.
    – Criggie
    Mar 15, 2020 at 11:15
  • All the devices are under the same network.. (192.168.0.x)
    – yaylitzis
    Mar 16, 2020 at 6:24

1) You are not routing. The router is irrelevant and distracting. Your B network is really like this:

enter image description here

All the direct traffic between camera and cctv device will be switched by the bottom switch only. Good

Traffic from a PC to/from the CCTV box will pass through three switches each way. Okay but not optimal.

Traffic between PC and NAS will stay in the top switch - good

Traffic to/from the internet will pass through two switches (the middle one being internal to the box you call a router) This is unavoidable really.

The only change I'd make would be to pick your best switch and use that as the top switch, call it "the core switch" and have a link out to each of the other two switches. I doubt your router's switch supports trunking or bonds, and may not support vlans.

The only traffic that will go over the entire network is broadcast traffic like ARP. And for a small enough network like this, will be fine. If you had a thousand devices on the same network, broadcast traffic may be an issue.

Depending on the needs of your organisation, there are some small downsides:

  • any PC can go and open each camera's web interface and possibly see the imagery. Same for the CCTV thing, so implement passwords.

  • If you have guest devices like visitor cellphones or tablets or laptops etc, then they can see everything in your flat network. At that time you will need to consider a separate network (vlan and SSID) for them.

As for your question about bandwidth, here's my linux host running motion for 8 cameras of various ages, but with four 1280x1024 and four at 1920x1080. The time window is 48 hours.

enter image description here
Yes the hostname is donkey, no I don't remember why.

The flatline is the overnight low - where each camera is in IR mode and sending a lower bitrate. During the day, they're seeing more colour, and more movement.

On that basis, a peak of 4.32 Mbit/sec, that's 540 kbits/second/camera. Ballpark 600 kbits/sec and your 16 cameras to the DVR will use less than 10 Mbit combined at full noise.

Of course if you have 10 PC computers all watching 10 cameras through the DVR at once, then this number grows pretty quick, 10 *10Mbit (outbound) + 10 Mbit (inbound) is 110 Mbit/sec which is roughly 11% of what a gigabit port can theoretically do, probably more like 15-20%.

So you have quite a lot of capacity for cameras.


I run setup a on an 60 camera setup. I still have the camera and recorder in a different vlan then client computers. The main thing I would recommend from a security perspective is use segmentation of your network.

We used to have setup B in our warehouse but needing some many extra switches and uplinks just is not necessary since if you monitor the interface loads there's not much data as you think.

This is the server uplink running 60 camera to milestone server

OK - [HP Ethernet 1Gb 4-port 331i Adapter] (Connected) 1 Gbit/s, In: 8.44 MB/s (7.1%), Out: 5.44 kB/s (0.0%)

enter image description here

  • Concur - I've got my DVR device with two (virtual) NICs so the cameras are on one VLAN with no access to anything, and all access to them is through what the DVR VM can see. Useful considering how dodgy cheap cameras like to phone-home to china.
    – Criggie
    Mar 16, 2020 at 6:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.