Timeline for How to connect 1 million devices over Layer 3 IP-protocol
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 29, 2018 at 19:19 | comment | added | aventurin | Consider using an IPv6 stack with support for SLAAC. However, given a million devices I would expect several failures per hour for the iot devices alone. Similar for the network hardware. I wonder if it wouldn't be more efficient and cost effective to simulate the whole thing in software. | |
| Jun 27, 2018 at 11:41 | answer | added | Tensibai | timeline score: 3 | |
| Jun 26, 2018 at 4:00 | history | edited | Ventures Joe | CC BY-SA 4.0 |
deleted 316 characters in body
|
| Jun 24, 2018 at 18:51 | comment | added | Ventures Joe | @hardillb Yes I can ignore layer 1 & 2 if I connect everything on PCB. Layer 1 and 2 are the Link layer - i.e. they put TCP/IP on a medium (e.g. Etherneet cable, Wi-Fi, fiber, whatever). I am happy to have everything on PCB copper. This is why some MCU have the MAC-layer functionality, but not the PHY one - so that you can provide the PHY IC that you need (Ethernet, WiFi, fiber, etc). I am trying to understand how to network together multiple IOT devices that are already TCP/IP-enabled, without being forced to a medium that I don't need, like Ethernet, WiFi, and so on. | |
| Jun 24, 2018 at 18:47 | comment | added | Ventures Joe | @JDAllen as for the "16 Class B subnets" or "1 Class A subnet", I don't understand whether you mean that this is an issue. The network will be self-contained inside a building. Is this a problem? I am happy to use switches/routers as necessary - but not off-the-shelf. I prefer to use ICs built on PCB to provide switching/routing as necessary. This is the information I am looking for. | |
| Jun 24, 2018 at 18:45 | comment | added | Ventures Joe | @JDAllen, TCP/IP does not require Ethernet (which is a layer 2 and 1. TCP/IP runs on anything (e.g. WiFi). So the information is not conflicting. As for the reference to USB, I clarified it is in reference to the size (as big as a USB stick) and to the physical connector (USB-connector, microSD connector, etc) - this is different from the USB-protocol (again, a layer 2/1 protocol). You did not understand the question correctly. Finally, I need TCP/IP because it is good network technology (for layer 3 and 4). You can suggest a better one if you know one. I think TCP/IP is good enough. | |
| Jun 24, 2018 at 11:28 | comment | added | hardillb | You can't just ignore Layer 1 & 2, you still need something for the IP to be carried over (especially if you have some sort of switch and want to do DHCP). | |
| Jun 24, 2018 at 0:21 | comment | added | JD Allen | Keep in mind too, that 1 million devices would require at least 16 Class B subnets (or 1 Class A subnet, like 10.x.x.x/8) with a number of hubs/routers required to do this right. | |
| Jun 24, 2018 at 0:11 | comment | added | JD Allen | Being a consultant, I try and get the clients to focus on the desired end state. You mention that you don't want Ethernet on all million devices, but then talk about putting an TCP/IP stack on every device...so conflicting information. You talk about connecting a number of devices to a 'board' using something like USB, so why do you need TCP/IP at all?? Where is the data from the IoT devices going? What kind of data is coming from the IoT device? Is it data that can be aggregated or averaged so that you could 'proxy' sets of IoT sensors as just one data point? | |
| Jun 23, 2018 at 21:39 | review | First posts | |||
| Jun 24, 2018 at 0:50 | |||||
| Jun 23, 2018 at 21:37 | history | asked | Ventures Joe | CC BY-SA 4.0 |