The Routing policy for a record defines how to answer a DNS query. Each type of policy does something different.  

The Simple routing policy provides the IP address associated with a name. With Simple routing an A record is associated with one or more IP addresses. A random selection will choose which IP to use.  It is important to note that Simple Routing policies do not support health checks. All other routing policies do. 

The Weighted routing policy is similar to simple routing and you can define a weight per IP address. Basically, you create records that have the same name and type and assign each record a numerical value that favors one IP address over another.  A value of 0 suggests a record is never returned. This is useful for simple load distribution or testing new software.  Each record is returned based on the weight compared to the total weight of all records. If a chosen record is Unhealthy, the process is repeated until a healthy record is obtained. 

The Geolocation routing policy tags records with a location that can be Default, Continent or Country. It allows you to distribute the IP of a resource that can cater to customers in different countries or different languages. It can also help you protect distribution or licensing rights. You can create a default record for IP addresses that do not map to a geographic location.  With geolocation routing an IP check verifies the customer’s location and the corresponding record for that location is returned based on the Location Tag for country, continent, or default.  

The Geo-proximity routing policy requires that you use Route 53’s traffic Flow feature and create a Traffic Policy.  A traffic policy is a resource that combines one or more routing policies.   Geo-proximity records are tagged with an AWS Region or using latitude and longitude coordinates.  Geo-proximity routing is based on distance and a defined bias. You can specify a Bias from -99 to 99. This is a value that you can use to route more traffic to an endpoint by using a positive value or Route less traffic to an endpoint by using a negative value. Use the bias of -99 to route the least amount of traffic to an endpoint.  You can think of the bias as being able to increase or decrease a region size in terms of coverage. This allows you to shift traffic from one location to another and route traffic based on the location of your resources. 

The Failover routing policy is able to route traffic to a primary resource and based on a health check re-direct traffic to a secondary resource. The re-direction happens if the health check fails.  Using failover routing you define a record to be primary and a different record to be secondary. You are also required to have a health check pre-defined. The routing of the primary record is active when the health check result is healthy.  Otherwise, the secondary record is used. 

The Latency routing policy chooses the record with the lowest latency to the customer.  You define multiple records with the same name and assign a region to each record.  AWS maintains a database of latency between the general location of users and the regions tagged in DNS records.  The record used is the one with the lowest recorded latency and is healthy.  This may not always be the closest resource, especially if the closest resource is saturated.

The Multi value Answer routing policy returns multiple IP addresses to a query. Up to 8 IP addresses corresponding to healthy records based on a health check are returned. If there are eight or less healthy hosts the response includes all healthy hosts.