performance testing of GB and GC

[tomkralidis]

Given various discussions on WTH, it would be valuable to put forth a testbed / experiment help measure performance of GBs and GCs with:

• deep topic levels
• large data files
• multiple MQTT publishers/subscribers/downloaders

cc @golfvert @kaiwirt @Haddouch101 @hhaddouch @efucile

[golfvert]

We are planned (almost agreed) to have hackathon week(s) next year to stress the system.
For that, we would need to define the scenario.

• Topic level depth is not an issue (up to a reasonable limit ~50 and we are far from it)
• Large files we already have the cache / no cache option. Do we need more ? Typically if a WIS2 Node proposes a large file ( what large is ?) then the GC MAY decide not to cache it ?
• Multiple MQTT: what is "multiple" ? 100s 1000's More ?

I would be more interesting in checking the reliability, the redundancy of the design.

There are also existing stress test of MQTT brokers.

Last but not least, we have only ONE publisher for a WIS2 Node and we have (in Météo-France design) one publisher per WIS2 Node. So, in any case, the number is small.