I think you have to hand it to Google, that is really next level deep analytics. I feel it would be hard for most companies without huge investment to do this (or maybe not).
It is like the reverse of a "customer survey" that instead of asking and getting an arbitrary number, instead of really detailed level of service performance.
People often consider to use this information for reliability and performance, but you can do much more with the data. For example, if a method has low latency, you can use short deadline with fast retry to improve reliability. If you see a sudden jump of certain usage, you can consider to use batching and caching to reduce your cost. If you see an unexpected usage of a service, you know someone introduce a new dependency in your system. Google teams often use the same data to understand how large services work and how they are correlated.
Comments
I think you have to hand it to Google, that is really next level deep analytics. I feel it would be hard for most companies without huge investment to do this (or maybe not).
It is like the reverse of a "customer survey" that instead of asking and getting an arbitrary number, instead of really detailed level of service performance.
People often consider to use this information for reliability and performance, but you can do much more with the data. For example, if a method has low latency, you can use short deadline with fast retry to improve reliability. If you see a sudden jump of certain usage, you can consider to use batching and caching to reduce your cost. If you see an unexpected usage of a service, you know someone introduce a new dependency in your system. Google teams often use the same data to understand how large services work and how they are correlated.
Disclosure: I worked on this feature at Google.