Hey you all!
I can try/will share some insights into whats going on and what is not.
First of all, i can definitely say, that the thread title is NOT correct. Both europe-north-1 and saltlake are hosted the exact same way. They even match all config values of those our customer opened satellites are configured with.
It might seem very strange, that europe-north-1 does not have any (or at least barely any) repair traffic. But as many previous replies stated, its a coincidence of a couple of factors:
-
The data that currently gets uploaded and has been uploaded via europe-north-1 is younger than every file hold by the saltlake satellite. This simply means, that data of either of the customer registration opened satellites or saltlake have to sustain the network influences much longer than the one on europe-north-1.
-
We have many nodes joining and some leaving the network every day. Especially the leaving part as significantly decreased, especially those that just leave without a graceful exit. This fact alone is a pretty huge influence into the amount of repair work ahead of a given satellite.
-
The repair threshold has been lifted to 52 pieces, causing much more frequent repair across the board to get better metrics around the velocity and affect of the above mentioned node churn and the impact of ongoing repair across the network. Lets not forget that per default all satellites share the nodes at the current stage (besides nodes that partially left some satellites).
All those points lead to a typical scenario, that the repair work tends to build up fairly quickly (as with the current upload speed to a given satellite, many segments reach that state at once) and then creates, the at this stage seemingly, “normal” repair throughput, that eg. saltlake is handling. We do expect to get into that zone on the europe-north-1 satellite soon, according to my rough calculations/estimations by end of this month.
Furthermore many of the tests we do against europe-north-1 and saltlake are mostly to test the internals of the satellite, under different load scenarios. As a sideeffect, we definitely upload huge amounts of data, but as all of this is frequently accessed and paid, it is a WIN-WIN situation for everyone, right?
TLDR: The data on europe-north-1 is pretty new/fresh compared to the data on the other satellites and especially lower node churn (without graceful exit) helps to extend that timeframe without any repair. In addition to our technically not needed high repair threshold, the repair load is intentionally higher than planned long term. We do not run any space reservation uploads or plan on doing those.
PS: As the operator of satellite.stefan-benten.de, i can assure that we have very exciting plans for the future of it. Sadly that’s all i can share so far. So, stay tuned!