Creating a wireless mesh from scratch

Last week, I had a (telephone) interview with a company who said that (they 
thought) their problem required a wireless mess solution but that the 
environment that they worked in was too noisy for the 2.4 GHz band that 
available solutions (such as Zigbee) operate in, and they wanted someone to 
design (eventually, code and test) such a solution using a different 
wireless band.

Not having any experience at all of working on a mesh, I declined on the 
grounds of insufficient match for the job.

My gut feel for this task is that creating (the software for) a mesh network 
from scratch would be 5-10 (man) year's of effort and not the 9-12 months 
that the company would like it to be, and I would be on a hiding to nothing 
trying.

Does anybody have any experience of creating a, software, mesh solution from 
scratch who can confirm or refute my estimate of how long it might take - 
should the opportunity arise again :-)

TIA

tim

 

Am 03.12.2014 um 23:36 schrieb tim.....:
> Last week, I had a (telephone) interview with a company who said that
> (they thought) their problem required a wireless mess solution
                                                    ^^^^
Someone just called, ranting about about some kind of slip of his he'll 
never hear the end of; does the name "Freud" ring any bell?

;-]

> but that
> the environment that they worked in was too noisy for the 2.4 GHz band
> that available solutions (such as Zigbee) operate in,

I can see how an environment might be too noisy "in", "on", or maybe 
"all across" a given frequency band, but "for" it?  Sounds like 
"customer is basing requirements on hearsay" to me.  Not to mention that 
the restriction to the 2.4 GHz band feels somewhat arbitrary.

tim..... wrote:
> Last week, I had a (telephone) interview with a company who said that
> (they thought) their problem required a wireless mess solution


They need to think again, then. Are they building to deploy MAN
networks? My understanding is that that market is oh so very dead.

> but that
> the environment that they worked in was too noisy for the 2.4 GHz band
> that available solutions (such as Zigbee) operate in,

Mesh over Zigbee sounds bloody painful.

> and they wanted
> someone to design (eventually, code and test) such a solution using a
> different wireless band.
>

These are available as COTS for next to nothing on 802.11g and such.
AliExpress has a node for $45.

I know it's comp.arch.embedded but Ethernet is a thing now, guys.

> Not having any experience at all of working on a mesh, I declined on the
> grounds of insufficient match for the job.
>

Mesh isn't that hard - it's just an extension of a bridge. There are 
things called "mesh bridges" but I have no idea what those are.

> My gut feel for this task is that creating (the software for) a mesh
> network from scratch would be 5-10 (man) year's of effort and not the
> 9-12 months that the company would like it to be, and I would be on a
> hiding to nothing trying.
>

Might be. Might not be.

> Does anybody have any experience of creating a, software, mesh solution
> from scratch who can confirm or refute my estimate of how long it might
> take - should the opportunity arise again :-)
>

No, but I worked on the firmware the original mesh.com guys put together 
well after Moto bought 'em.

> TIA
>
> tim
>
>
>
-- 
Les Cargill

Hello Tim,

"tim....." <tims_new_home@yahoo.co.uk> wrote in
news:ce9hk4F1abpU1@mid.individual.net: 

> Last week, I had a (telephone) interview with a company who said that
> (they thought) their problem required a wireless mess solution but
> that the environment that they worked in was too noisy for the 2.4 GHz
> band that available solutions (such as Zigbee) operate in, and they
> wanted someone to design (eventually, code and test) such a solution
> using a different wireless band.

It sounds like they don't have the requirements nailed down yet.  Any 
single candidate won't be able to solve their issue in the time they 
want.

You could have inquired about the requirements for the project, what 
issues they were having, and how you might go about solving their 
problem ("get a solid communications network for their product").
 
> Not having any experience at all of working on a mesh, I declined on
> the grounds of insufficient match for the job.

That was a wise decision.  However if you have a good grounding in 
communications applications you might be able to get up to speed on what 
is needed.

I've been in that situation too, and the time frame is what worried me 
most about your call.  They were looking for a savior as well as someone 
to solve a vague set of requirements.

With enough communication background or interest in their product you 
might have gone on to a second interview which might have led you to 
people who could have told you what the facts really were about their 
goals.

> My gut feel for this task is that creating (the software for) a mesh
> network from scratch would be 5-10 (man) year's of effort and not the
> 9-12 months that the company would like it to be, and I would be on a
> hiding to nothing trying.

Hopefully they'd build on an existing hardware/firmware/communications 
stack and have you, or your team, solve the problem from there.  9-12 
months isn't long for an outsider to make a great solution that would be 
production worthy.  They might take that long to evaluate possible 
communications technologies if they haven't already done so.
 
> Does anybody have any experience of creating a, software, mesh
> solution from scratch who can confirm or refute my estimate of how
> long it might take - should the opportunity arise again :-)

The length of time depends mostly on the complexity and richness of the 
product.  Suppose you have radios and stacks that provide the ability to 
monitor other stations in your area and a store/forward communication 
protocol.  A simple application could take relatively short time.  A 
more complex application would have more problems to solve to evolve a 
solid solution.  Again, this presumes they really do need a mesh 
network.
 
> TIA
> 
> tim

Good luck,

David

Den torsdag den 4. december 2014 01.10.43 UTC+1 skrev Les Cargill:
> tim..... wrote:
> > Last week, I had a (telephone) interview with a company who said that
> > (they thought) their problem required a wireless mess solution
> 
> 
> They need to think again, then. Are they building to deploy MAN
> networks? My understanding is that that market is oh so very dead.
> 
> > but that
> > the environment that they worked in was too noisy for the 2.4 GHz band
> > that available solutions (such as Zigbee) operate in,
> 
> Mesh over Zigbee sounds bloody painful.
> 

it's been part of the spec since the beginning, though I don't know how 
well it works, it's probably been 8 years since I worked on zigbee


> > and they wanted
> > someone to design (eventually, code and test) such a solution using a
> > different wireless band.
> >
> 
> These are available as COTS for next to nothing on 802.11g and such.
> AliExpress has a node for $45.
> 

these are ~$5 in onesies 
https://nurdspace.nl/ESP8266

but I don't see wifi being more robust in noise than Zigbee

-Lasse

On 12/3/2014 3:36 PM, tim..... wrote:
> Last week, I had a (telephone) interview with a company who said that (they
> thought) their problem required a wireless mess solution but that the
> environment that they worked in was too noisy for the 2.4 GHz band that
> available solutions (such as Zigbee) operate in, and they wanted someone to
> design (eventually, code and test) such a solution using a different wireless
> band.

"They thought"... what?  That they needed a wireless mesh network?  That
the RF environment was too polluted (at least around 2.4G)?

Did they rule out a *wired* network?  What sort of traffic were they
looking at?  Were nodes mobile or stationary?  How robust did the
mesh need to be?  etc.

> Not having any experience at all of working on a mesh, I declined on the
> grounds of insufficient match for the job.

I'd decline based on "insufficient detail"!  :-/

> My gut feel for this task is that creating (the software for) a mesh network
> from scratch would be 5-10 (man) year's of effort and not the 9-12 months that
> the company would like it to be, and I would be on a hiding to nothing trying.
>
> Does anybody have any experience of creating a, software, mesh solution from
> scratch who can confirm or refute my estimate of how long it might take -
> should the opportunity arise again :-)

There are just too many details missing, here, to even begin to see what
their *real* requirements are likely to be.  There's a huge difference between
a general purpose, dynamically reconfigurable mesh and, for example, a very
*static* mesh.  Likewise, data rates, error protocols, etc.:  "How much
of the data *must* get through?  What guarantees can you provide as to the
cleanliness of the spectrum in <whatever> band selected?"

On Wed, 3 Dec 2014 22:36:52 -0000, "tim....."
<tims_new_home@yahoo.co.uk> wrote:

>Last week, I had a (telephone) interview with a company who said that (they 
>thought) their problem required a wireless mess solution but that the 
>environment that they worked in was too noisy for the 2.4 GHz band that 
>available solutions (such as Zigbee) operate in, and they wanted someone to 
>design (eventually, code and test) such a solution using a different 
>wireless band.

A mesh network may be attractive in very low duty cycle applications,
say each node sends some measurements for a few hundred milliseconds
every minute, but in a larger network based on half duplex single
frequency store and forward using CSMA band access, will quickly
become jammed. 

This is especially the case, if the stations are too close together
that they also hear other stations beyond the nearest few neighbors.

Perhaps "too noisy" refers to the interference created by their
_own_network_nodes operated geographically at a larger distance, than
the nearest nodes used for actual mesh communication.

If this mesh is connected to the external world, the stations closest
to the gateway node will quickly consume the whole RF bandwidth.
 

langwadt@fonz.dk wrote:
> Den torsdag den 4. december 2014 01.10.43 UTC+1 skrev Les Cargill:
>> tim..... wrote:
>>> Last week, I had a (telephone) interview with a company who said that
>>> (they thought) their problem required a wireless mess solution
>>
>>
>> They need to think again, then. Are they building to deploy MAN
>> networks? My understanding is that that market is oh so very dead.
>>
>>> but that
>>> the environment that they worked in was too noisy for the 2.4 GHz band
>>> that available solutions (such as Zigbee) operate in,
>>
>> Mesh over Zigbee sounds bloody painful.
>>
>
> it's been part of the spec since the beginning, though I don't know how
> well it works, it's probably been 8 years since I worked on zigbee
>

You are absolutely correct -  I was thinking of something else.

>
>>> and they wanted
>>> someone to design (eventually, code and test) such a solution using a
>>> different wireless band.
>>>
>>
>> These are available as COTS for next to nothing on 802.11g and such.
>> AliExpress has a node for $45.
>>
>
> these are ~$5 in onesies
> https://nurdspace.nl/ESP8266
>
> but I don't see wifi being more robust in noise than Zigbee
>

No, I wouldn't think so.

> -Lasse
>
-- 
Les Cargill

tim..... wrote:

> Last week, I had a (telephone) interview with a company who said that
> (they thought) their problem required a wireless mess solution but that
> the environment that they worked in was too noisy for the 2.4 GHz band
> that available solutions (such as Zigbee) operate in, and they wanted
> someone to design (eventually, code and test) such a solution using a
> different wireless band.
> 
> Not having any experience at all of working on a mesh, I declined on the
> grounds of insufficient match for the job.
> 
> My gut feel for this task is that creating (the software for) a mesh
> network from scratch would be 5-10 (man) year's of effort and not the 9-12
> months that the company would like it to be, and I would be on a hiding to
> nothing trying.
> 
> Does anybody have any experience of creating a, software, mesh solution
> from scratch who can confirm or refute my estimate of how long it might
> take - should the opportunity arise again :-)

OLPC is supposed to self-organise a mesh when it finds itself in range of 
other nodes. However, you may rule it out on other grounds.

I would have stated that there should be a feasibility study done first, 
which would involve getting in a spectrum analyser and a sensitive wide band 
receiver to see what his wireless environment looked like (for real). You 
would, at least have been able to make a proper report of the environment 
anyone else would be working in and would have a good knowledge of whether 
any other solution might have suggested itself as you wandered around his 
plant.


-- 
********************************************************************
Paul E. Bennett IEng MIET.....<email://Paul_E.Bennett@topmail.co.uk>
Forth based HIDECS Consultancy.............<http://www.hidecs.co.uk>
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-510979
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************

David LaRue wrote:

> Hello Tim,
> 
> "tim....." <tims_new_home@yahoo.co.uk> wrote in
> news:ce9hk4F1abpU1@mid.individual.net:
> 
>> Last week, I had a (telephone) interview with a company who said that
>> (they thought) their problem required a wireless mess solution but
>> that the environment that they worked in was too noisy for the 2.4 GHz
>> band that available solutions (such as Zigbee) operate in, and they
>> wanted someone to design (eventually, code and test) such a solution
>> using a different wireless band.
> 
> It sounds like they don't have the requirements nailed down yet.  Any
> single candidate won't be able to solve their issue in the time they
> want.
> 
> You could have inquired about the requirements for the project, what
> issues they were having, and how you might go about solving their
> problem ("get a solid communications network for their product").
>  
>> Not having any experience at all of working on a mesh, I declined on
>> the grounds of insufficient match for the job.
> 
> That was a wise decision.  However if you have a good grounding in
> communications applications you might be able to get up to speed on what
> is needed.
> 
> I've been in that situation too, and the time frame is what worried me
> most about your call.  They were looking for a savior as well as someone
> to solve a vague set of requirements.
> 
> With enough communication background or interest in their product you
> might have gone on to a second interview which might have led you to
> people who could have told you what the facts really were about their
> goals.
> 
>> My gut feel for this task is that creating (the software for) a mesh
>> network from scratch would be 5-10 (man) year's of effort and not the
>> 9-12 months that the company would like it to be, and I would be on a
>> hiding to nothing trying.
> 
> Hopefully they'd build on an existing hardware/firmware/communications
> stack and have you, or your team, solve the problem from there.  9-12
> months isn't long for an outsider to make a great solution that would be
> production worthy.  They might take that long to evaluate possible
> communications technologies if they haven't already done so.
>  
>> Does anybody have any experience of creating a, software, mesh
>> solution from scratch who can confirm or refute my estimate of how
>> long it might take - should the opportunity arise again :-)
> 
> The length of time depends mostly on the complexity and richness of the
> product.  Suppose you have radios and stacks that provide the ability to
> monitor other stations in your area and a store/forward communication
> protocol.  A simple application could take relatively short time.  A
> more complex application would have more problems to solve to evolve a
> solid solution.  Again, this presumes they really do need a mesh
> network.

A Mesh network requires having enough residual bandwidth for each node that 
it can repeat frames intended for another node between two nodes not able to 
link directly.

Node 1 message meant for node 3 is received by node 2 then gets repeated to 
node 3 by node 2. 

-- 
********************************************************************
Paul E. Bennett IEng MIET.....<email://Paul_E.Bennett@topmail.co.uk>
Forth based HIDECS Consultancy.............<http://www.hidecs.co.uk>
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-510979
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************