 |
| Photo by Atypeek Dgn |
AWS Global Accelerator enables low latency networking for applications such as gaming or VOIP.
Anycast IP
- Broadcast
- Multicast
- Unicast
- Anycast
With Anycast, two or more nodes can share the same IPv6 address. The router will route packets to the nearest node. The diagram below illustrates this - even though two nodes have the same address, packets are routed to the closest node indicated in green. It is possible to use IPv4 addresses via a BGP workaround.
Accelerators
With Anycast IP, it is possible to have multiple AWS edge locations advertise the same IP address. As a result, users closer to an edge location are immediately directed to AWS Global Accelerator nodes.
The edge locations do not host the application servers. They provide an entry point to the AWS Global Accelerator that allows low latency transit to the application servers over the highly available and private AWS global backbone network. The network packets would have to otherwise traverse the internet.
Clients in Mumbai, for example, can access an application backend via a local accelerator and AWSs private global backbone network, as opposed to having to access it over the internet.
AWS Global Accelerator includes 2 static IP addresses. It is also possible to bring your own IP addresses.
Benefits
Jumbo Frames
AWS's network supports Jumbo Frames, which allow larger TCP packets. This reduces the ACK/SYN networking overhead and can significantly reduce latency.
Monitoring
Global Accelerators protect against DDOS attached via AWS Sheild
Failover
AWS's private network has a high degree of monitoring and supports redundancy and failover of internal components. Additionally, a globally distributed application will benefit by automatically having traffic rerouted in case of a region failover.
Types of Global Accelerators
Standard
The Standard Global Accelerator provides support for routing traffic to load balancers, elastic IP addresses, and EC2 instances.
Custom
The Custom Global Accelerator provides support for routing traffic to a specific destination and port in a private subnet in a VPC. This allows the user to define custom logic in handling network traffic.
