Are you selling caches? No, we're just showing

There are frequent requests in Dodo, and it happens that with a response body of several megabytes (for example, the menu for a mobile application weighs so much). With the growth of the user base (and the increase in the number of requests), the team faced a problem — the servers were no longer able to cope. The team didn't want to add new servers, and as a result, they sat down for caching.

Anton and Andrey want to tell:

• what paths of data caching went through (internal in-memory and distributed caches, caching responses on the application side, caching on nginx)
• what problems are with each of them
• how to make the most frequent request to the site's API work without caches at all;
• plans.

The talk will be interesting for those who create mass services with a large number of users.