AnycastWikipedia: Anycast is a network addressing and routing methodology in which datagrams from a single sender are routed to the topologically nearest node in a group of potential receivers, though it may be sent to..., at a very basic level, is when a collection of servers share the same IP address and data is sent from a source computer to the server that is topographically closest. It is important to remember that topographically closer does not inherently mean geographically closer, though this is often the case.
Anycast is used primarily for load balancingWikipedia: Load balancing is a computer networking method for distributing workloads across multiple computers or a computer cluster, network links, central processing units, disk drives, or other resources. Suc... to allow the server topographically closest to a user to handle their request. This helps cut down on latency and bandwidth costs and improves load time for users.
Anycast is linked with the Border Gateway ProtocolWikipedia: Border Gateway BC Protocol (BGP) is the protocol which is used to make core routing decisions on the Internet; it involves a table of IP networks or "prefixes" which designate network reachability amo.... This is a protocol used between routersWikipedia: A router is a device that forwards data packets between computer networks, creating an overlay internetwork. A router is connected to two or more data lines from different networks. When a data packet... on the Internet with the intent of ensuring that all of a router’s neighbours are aware of the networks that can be reached through that router and the topographical distance to those networks. The principal of Anycast is that a single IP address is advertised in the BGP messages of multiple routers. As this propagates across the Internet, routers become aware of which of their neighbours provides the short topographical path to the advertised IP address.
IP addresses used in Anycast are often purchased directly from a Regional Internet registryWikipedia: A regional Internet registry (RIR) is an organization that manages the allocation and registration of Internet number resources within a particular region of the world. Internet number resources inclu.... Some data centersWikipedia: A data center or computer centre (also datacenter) is a facility used to house computer systems and associated components, such as telecommunications and storage systems. It generally includes redunda... are known to rent IP addresses to customers and allow them to be advertised by other data centres.
As with all routing, it cannot be guaranteed that a packet will take the same path across the Internet as its predecessor. With Anycast, it cannot be guaranteed that a packet will reach the same destination server as its predecessor. As such, Anycast is not suitable for protocols which track state. TCP is an example of one of these. UDP, however, is perfect for Anycast providing it does not try to track state at a higher level of the OSI model and that the application layer protocol does not rely on a large number of fragemented datagrams to transfer data.
The typical scenario for Anycast as a load balancer is thus:
- A server in London has its own IP address 3.3.3.3 and a shared Anycast IP address 1.1.1.1.
- A server in New York has its own IP address 4.4.4.4 and a shared Anycast IP address 1.1.1.1.
- Each of the above servers runs a DNSWikipedia: The Domain Name System (DNS) is a hierarchical distributed naming system for computers, services, or any resource connected to the Internet or a private network. It associates various information with... server listening on 1.1.1.1.
- The DNS servers serve up an A record for anycastdomain.com. London would serve up 3.3.3.3 and New York would serve up 4.4.4.4.
- When a DNS request is made for anycastdomain.com, Anycast would route this request to its topographically closest DNS server. This DNS server would, in turn, serve up the unique IP address of its own server and a TCP connection would be established over standard unicast.
Feedback from companies such as ScaleEngine is that it’s quite difficult to persuade data centres to add IP addresses to their BGP. This appears to be best suited to larger organisations who lease their own transitWikipedia: Internet transit is the service of allowing network traffic to cross or "transit" a computer network, usually used to connect a smaller Internet service provider (ISP) to the larger Internet. Technica... and have BGP agreements with their transit providers.