Thread mesh network capacity problem

RE: Thread mesh network capacity problem

Rorisang — Thu, 08 Nov 2018 12:11:55 GMT

Ok. Thanks will do exactly that.

RE: Thread mesh network capacity problem

Hung Bui — Thu, 08 Nov 2018 12:10:15 GMT

@Rorisang: Please create a new case and describe in more detail what you want to achieve and the problem you are seeing.

RE: Thread mesh network capacity problem

Rorisang — Mon, 05 Nov 2018 13:25:43 GMT

Hi, Wojciech. Please excuse me but i need your help and that of everyone else following the thread. I am stuck trying to set up a similar mesh network but using just 9 nodes. How may i do that effectively? I need to also collect latency data for the broadcasts. Please i seriously need help.

RE: Thread mesh network capacity problem

Wojtek — Mon, 29 Jan 2018 09:21:28 GMT

Hi Piotr, Thanks for response! Library update will be first thing to try on my list. Best Regards, Wojtek

RE: Thread mesh network capacity problem

Piotr Szkotak — Thu, 25 Jan 2018 11:44:18 GMT

Hi Wojciech,

Thank you for the detailed answers. It seems that are indeed some issues with running bigger mesh networks with SDK 0.10.0. I was able to observe it as well.

As there were many bugfixes between 0.10.0 and 0.11.0 instead of troubleshooting an older release I performed the same test with a newer one. To speed up testing I flashed one node with the CLI for TTM and remaining ones with simple_coap_server as they join the network on their own. I flashed five kits, waited for 20s and flashed another five. To make sure I used over 30 kits.

Please see attached TTM screen.

Could you try 0.11.0 with your application?

Kind regards, Piotr Szkotak

RE: Thread mesh network capacity problem

Wojtek — Wed, 24 Jan 2018 13:41:30 GMT

Hi Piotr! My nodes are programmed with V0.10. My Topology Monitor is PDK with CLI example from V0.11. Battery powered nodes are hardcoded to work as SED but other ones can choose mode on their own. Most of network consist of routers. I know about limit of 32 of these. Im attaching nodes one after another starting from routers delegated to be the skeleton of the network. When all nodes are attached network works for a while until crash. When I power down about 3 nodes network starts to work properly. My previous experiences with Thread was that nodes need to be attached with a liitle time beetween each other to avoid problems. I've attached screen from TTM in first post. The network on it is on edge of stability.

RE: Thread mesh network capacity problem

Piotr Szkotak — Wed, 24 Jan 2018 13:03:26 GMT

Hi Wojtek,

Thank you for your interest in Thread SDK, your setup is pretty impressive. Let's try to confirm some basics before we start to dig deeper.

Could you confirm which SDK is used (0.10.0 or 0.11.0)?

How many routers and how many childs are in your setup? Please note that by default in OpenThread a single router supports up to 10 children. Library would need to be recompiled with changed OPENTHREAD_CONFIG_MAX_CHILDREN define.

If this is not your case is it possible for you to share screen from the TTM and wireshark capture?

Do you connect all nodes at once to the network?

Did you observe any correlation between order in which you power on nodes and the maximum number of nodes supported in your case?

Kind regards, Piotr Szkotak

RE: Thread mesh network capacity problem

Wojtek — Wed, 24 Jan 2018 12:05:02 GMT

_ 0> 0> app: Thread version: OPENTHREAD/ga89eb88-dirty; NRF52840; Nov 17 2017 09:01:26 0> 0> app: Network name: OpenThread 0> 0> app: State changed! Flags: 0x00001311 Current role: 0 0> 0> app: State changed! Flags: 0x00000100 Current role: 0 0> 0> app: State changed! Flags: 0x00000004 Current role: 1 0> 0> app: State changed! Flags: 0x00000104 Current role: 1 0> 0> app: State changed! Flags: 0x00000004 Current role: 1 0> 0> app: State changed! Flags: 0x00000004 Current role: 1

It spawns this info message infinietely. Device is programmed with CLI example from sdk v0.11 .

RE: Thread mesh network capacity problem

Wojtek — Wed, 24 Jan 2018 10:44:48 GMT

Hi, Application sends a packet of data from sensors every 1 minute by COAP. In case of lack of connection messages are buffered and sent while connection renew. When problem occurs packets from network are no longer forwarded to server and border router cannot join network (strange led blinking). Big amount of traffic can occur while many nodes empty their message buffers. This happens very often propably because network unstability. I can see in Topology Monitor that network rebuilds all the time. I will post RTT log as soon I set a new test.

RE: Thread mesh network capacity problem

Hung Bui — Wed, 24 Jan 2018 10:12:20 GMT

Hi Wojtek,

Could you tell about the application you run on the network ? was there lots of traffic between nodes ?

When the issue happened, can other nodes in the network communicate to each other ?

What do you have in the log of the device that can't join the network ?