InAround the next month?,end of March, traffic started being routed through our new data center in Singapore (cf. traffic will start being routed through our new data center in SingaporeT189252 and [[https://github.com/wikimedia/operations-dns/commits/master/geo-maps |detailed timeline by country]]). The performance team is going to be measuring the hell out of performance impacts for the users this impactshas since [[https://blog.wikimedia.org/2018/04/24/new-data-center-singapore/ | published data]] on the performance impacts for readers, and this is a wonderful opportunity to test the hypothesis that if you make something faster on the internet, people use it more.
Using pageviews and/or user frequency (@Tbayer will have details on this metric) let's look at any impact we can determine as a result of the change.Let's look at any impact we can determine as a result of the change, using daily time series for the following metrics:
- Retention (average return time within 31 days based on the last-access data recorded in `tbayer.webrequest_extract_bak`)
We will need:For the following countries/sets of countries:
- a list of the geographies this will impact (or some other way to distinguish the relevant traffic)- Bangladesh
- important transition dates- Japan
- results of the performance team's analysis (for possible correlation)
(Exclude Singapore itself per T184677#4188563 ff.)