You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It will be more optimal - and of course it is prelude to Change Data Capture technique - to add an implementation of schema versioning with indexes instead of repeated metrics names like this:
as you can see, we have a bigger S item code/kind for schema definition and all repeated metrics names are indexed by hexadecimal 3 positional digits. It gives us ~4k metrics in one row/message.
With schema + versioning, we can save approx. 50% of message size. Much more, compared to JSON formated the same message!
In general with the schema-less basic version of the protocol we can send data like this:
It will be more optimal - and of course it is prelude to Change Data Capture technique - to add an implementation of schema versioning with indexes instead of repeated metrics names like this:
as you can see, we have a special item
s|v1
- information about versioning and version name.At the same time schema message needs to be sent to the server/receiver, too - schema format proposal:
as you can see, we have a bigger
S
item code/kind for schema definition and all repeated metrics names are indexed by hexadecimal 3 positional digits. It gives us ~4k metrics in one row/message.With schema + versioning, we can save approx. 50% of message size. Much more, compared to JSON formated the same message!
The text was updated successfully, but these errors were encountered: