[Legba2013] PapaLegba 2013 review.

[tim p. <th...@we...>]

I've just spent the day working on the stats and I thought I'd write a note on
my thoughts for the PapaLegba crew and my Tropo sponsors.

http://gont.westhawk.co.uk/bm2013/stats/report.htm

We did significantly worse than last year. Fewer calls, fewer minutes, fewer hours up.

We did provide a useful service - e.g. 80 5 min long calls that wouldn't have been made any other way.

We also managed to gather some raw data to analyse - which is helpful - thanks Mike!

What went wrong?

1) We set up late. by the time we went live there were already lots of handsets to register.
2) The SMS delivery mechanism lacked the throughput to deliver the registration load, leaving no capacity for normal traffic. Even when it had been re-written (twice) on the Playa it barely kept up. When I added 2000 sms's notifying of rain on Monday we managed to only deliver 61 of them by the time we tore down.
3) We lost power _every_ night. The meant that any phone still powered up had to re-register every morning - even if it hadn't moved. This consumed much of our radio capacity.
4) We had too few towers.
5) The design kept changing - despite the promise that we'd test everything at RELIEF, use a VPN and use IMSIs as caller IDs - we did none of the above. 
We rocked up late, with an entirely untested core network which was still being re-written on thursday. It is frankly amazing that anything worked at all.
6) We broke our promise to Ice to deliver a gate SMS application through to monday evening.
7) we didn't communicate our purpose very well to other burners.

What worked?
1) It looks like the new L3 held up ok, although I have no stats on that.
2) The ability to script the core network meant that we could re-write the SMS delivery (at least to some extent) so that it prioritized recent SMSs thus keeping a small number of people happy at least.
3) the interface with tropo worked - the ability to access a database from tropo was very handy for sms groups etc.
4) outbound voice and SMS to the PSTN
5) The inter-base station 3.5GHz network
6) PlayaNet as backhaul worked ok for us this year - perhaps becuase we didn't require anything special of it.
7) The provision of wifi and a hardphone was popular again.
8) group sms


Overall, I'm doubtful I'll do this again, it will be difficult to justify the time spent for the results unless we make significant changes in our approach.

I'll write a separate note on the viability of scripting a core network in javascript.

Tim.