Re: [manet] (DLEP) Relative Link Quality and routing metrics

Assume two types of radio's.

First one is 802.11 type of radio. It has rate adaption and retry statistics. It reports CDR, this is the *good-put* the radio would have taking all the airtime. In this case, utility of RLQ is low.
But when the radio reports only the average or most recent *TX rate*, the RLQ can be of value. At higher rates, RLQ would be 100, because rate adaption takes care of variation in link budget. But at lower rates and high retry ration (and backoff timers... RLQ would provide kind of fast DAT, when RLQ is based on RSSI and/or retry counters.

Second one is more like a legacy data-radio. It has fixed rate, fixed TX power and it is often used for broadcast, i.e. the messages are addressed to all receivers. For the sender, RLQ is always 100, it doesn't mater if there is any receiver. In military, it is called fire and forget. But for receivers, there is a variation in metrics. Nearby nodes would receive all packets, nodes far from the sender would receive nothing. Somewhere between, nodes would receive part of the messages sent by a node. This is called the grey zone. Reported metrics could be loss ratio, RSSI, Eb/No or RLQ. Here, RLQ could be a compound value, based on the mentioned metrics.

Because RLQ would be very specific to the type of radio, you could design some kind of plugin or profile mechanism, where an administrator could select a calculation model.

I did some work using RSSI to RLQ mapping, the router used that to calculate the routing protocol metric. I used a lookup table. Tuning the table is tough, different conditions need different mapping.

Teco

> Op 20 apr. 2018, om 13:05 heeft Henning Rogge <hrogge@gmail.com> het volgende geschreven:
> 
> Sorry,
> 
> but this does not help...
> 
> lets say I estimate the link cost for a link (based on data-rate) to be 1000...
> 
> if the radio reports RLQ=100, I would keep the 1000... but how do I
> modify it for a RLQ of 80... or 50... or 1?
> 
> If I cannot integrate the value into the cost metric, I have no use
> for the value. Unfortunately it is often the ONLY value the radio
> reports.
> 
> Henning
> 
> On Fri, Apr 20, 2018 at 1:02 PM, Rick Taylor
> <rick@tropicalstormsoftware.com> wrote:
>> I have always suggested that RLQ should be a measure of how hard the
>> modem is working to maintain the link, i.e. the higher the RLQ, the
>> more stable the link is, and a low RLQ indicates that the link may well
>> radically change metrics, or dissapear soon.
>> 
>> One can imagine a sophisticated modem maintiaining a link in very
>> adverse conditions, reporting a low RLQ, but otherwise good metrics.
>> 
>> Whether RLQ is a good metric to use as a route cost is a more difficult
>> question.  I've always been of the opinion that RLQ and Resources make
>> a good tie-breakers, but CDR and Latency make better 'primary' metrics.
>> 
>> Hope that helps a little?
>> 
>> Rick
>> 
>> On Fri, 2018-04-20 at 10:00 +0200, Henning Rogge wrote:
>>> Hello,
>>> 
>>> I am currently looking for a good way to integrate the RLQ value of
>>> DLEP into a cost based routing metric, e.g. DAT. But I am not sure
>>> how
>>> to do this...
>>> 
>>> has anyone here good experience using RLQ and maybe an advise how
>>> "hard" you should penalize a link with a RLQ less than 100?
>>> 
>>> Henning Rogge
>>> 
>>> _______________________________________________
>>> manet mailing list
>>> manet@ietf.org
>>> https://www.ietf.org/mailman/listinfo/manet
> 
> _______________________________________________
> manet mailing list
> manet@ietf.org
> https://www.ietf.org/mailman/listinfo/manet