Cloudfront decision to fetch data from cache or server?

Question

From Amazon cloud front

Amazon CloudFront is a web service that speeds up distribution of your static and dynamic web content, such as .html, .css, .php, and image files, to your users. CloudFront delivers your content through a worldwide network of data centers called edge locations.

Per mine undserstanding, CloudFront must be caching the content with URL as key. URL can serve both static and dynamic content. Say i have 100 weburl's , out of which 30 serves the static content and 70 serves dynamic content(user specific data). I have one question each on static and dynamic content

Dynamic content :- Say user_A access his data through url_A from US. That data has been cached . He updates the first name. Now same user will access the data from same location in US or from another location in UK. We he see data prior to updation. If yes how will edge location come to know data needs to fetched from server not from cache ? Does edge location continue to display the data from cache for configurable amount of time and if time is passed then fetch it from server ?

Does cloudfront allows to configure specific URL's that needs to fetched from server instead of cache always ?

Static content :- There are chances that even static data may change will with each release. How cloud front will know that cached static content is stale and needs to be fetched from server ?

Answer 1

Amazon CloudFront uses an expiration period (or Time To Live - TTL) that you specify.

For static content, you can set the default TTL for the distribution or you can specify the TTL as part of the headers. When the TTL has expired, the CloudFront edge location will check to see whether the Last Modified timestamp on the object has changed. If it has changed, it will fetch the updated copy. If it is not changed, it will continue serving the existing copy for the new time period.

If static content has changed, your application must send an Invalidation Request to tell CloudFront to reload the object even when the TTL has not expired.

For dynamic content, your application will normally specify zero as the TTL. Thus, that URL will always be fetched from the origin, allowing the server to modify the content for the user.

A half-and-half method is to use parameters (eg xx.cloudfront.net/info.html?user=foo). When configuring the CloudFront distribution, you can specify whether a different parameter (eg user=fred) should be treated as a separate object or whether it should be ignored.

Also, please note that each CloudFront edge location has its own cache. So, if somebody accessed a page from the USA, that would not cause it to be cached in the UK.

See the documentation: Specifying How Long Objects Stay in a CloudFront Edge Cache (Expiration)