Global load balancing on a content delivery network
View Patent ↗The invention relates to systems and methods of global load balancing in a content delivery network having a plurality of edge servers which may be distributed across multiple geographic locations. According to one aspect of the invention, a global load balancing system includes a first load balancing server for receiving a packet requesting content to be delivered to a client, selecting one of the plurality of edge servers to deliver the requested content to the client, and forwarding the packet across a network connection to a second load balancing server, which forwards the packet to the selected edge server. The selected edge server, in response to receiving the packet, sends across a network connection the requested content with an address for direct delivery to the client, thereby allowing the requested content to be delivered to the client while bypassing a return path through the first load balancing server.
1. A content delivery network (CDN) comprising:
a first Point-of-Presence (POP) comprising a first load balancing server and a first set of edge servers, the first load balancing server assigned a particular Internet protocol (IP) address such that a request for content addressed to the particular IP address originating from a first region that is proximate to the first POP is routed to the first load balancing server; and
a second POP comprising a second load balancing server and a second set of edge servers, the second load balancing server assigned the same particular IP address as the first load balancing server such that a request for the content addressed to the particular IP address originating from a second region that is proximate to the second POP is routed to the second load balancing server, the second load balancing server comprising:
a processor (i) processing a set of parameters to identify whether the first POP offers improved delivery of content relative to the second POP or the second POP provides improved delivery of the content relative to the first POP and (ii) selecting an edge server of the first set of edge servers when the processor determines that the first POP provides improved delivery for the content relative to the second POP based on the set of parameters; and a packet switch forwarding a received request for the content to the first POP when the processor determines that the first POP provides improved delivery for the content relative to the second POP based on the set of parameters.
2. The CDN of claim 1 , wherein the packet switch forwards the received request from the second load balancing server to the selected edge server of the first set of edge servers.
3. The CDN of claim 1 , wherein the packet switch forwards the received request from the second load balancing server to the first load balancing server for subsequent forwarding of the request from the first load balancing server to the selected edge server of the first set of edge servers.
4. The CDN of claim 1 , wherein the packet switch is further for forwarding the received request for the content to an edge server selected from the second set of edge servers when the processor determines that the second POP provides improved delivery for the content relative to the first POP based on the set of parameters.
5. The CDN of claim 1 , wherein the second load balancing server further comprises an input port receiving the request, the request originating from the second region and addressed to the particular IP address.
6. The CDN of claim 1 , wherein the set of parameters comprises cost of delivering content from an edge server of at least the second set of edge servers to a recipient that originates the request.
7. The CDN of claim 1 , wherein the set of parameters comprises at least one of server load, server availability, and network latency for an edge server of at least the second set of edge servers.
8. The CDN of claim 1 , wherein the packet switch forwards the request with an address, the address for delivering the content directly to a requestor that originates the request while bypassing a return path through any of the first and second load balancing servers.
9. The CDN of claim 1 further comprising (i) a first router associated with the first POP, the first router advertising the particular IP address to at least one other router that is proximate to the first POP and (ii) a second router associated with the second POP, the second router advertising the particular IP address to at least one other router that is proximate to the second POP.
10. A method of load balancing across a content delivery network (CDN), the CDN comprising a first Point-of-Presence (POP) and a second POP from which content is distributed on behalf of others by the CDN, the first POP geographically separated from the second POP, the method comprising:
configuring the CDN for Anycast routing by (i) assigning a particular IP address to the first POP and (ii) assigning the same particular IP address to the second POP, wherein said assigning the particular IP address to the first POP facilitates routing to the first POP, a request for content that is addressed to the particular IP address and that originates from a first region that is proximate to the first POP, and wherein said assigning the particular IP address to the second POP facilitates routing to the second POP, a request for the content that is addressed to the particular IP address and that originates from a second region that is proximate to the second POP;
monitoring a set of parameters relating to performance of at least the first POP and the second POP;
receiving at the first POP, a request for content that is addressed to the particular IP address and that originates from the first region;
determining by the first POP, based on the set of parameters whether the second POP provides improved delivery of the content relative to the first POP; and
selecting by the first POP, an edge server operating in the second POP when the second POP provides improved delivery of the content relative to the first POP: and
forwarding the request to the second POP when the second POP provides improved delivery of the content relative the first POP.
11. The method of claim 10 further comprising selecting an edge server operating in the first POP and forwarding the request to the edge server when the first POP provides improved delivery of the content relative to the second POP.
12. The method of claim 10 , wherein the selected edge server, in response to receiving the request, sends the content to a requestor that originates the request.
13. The method of claim 10 , wherein determining based on the set of parameters comprises determining based on at least one of cost of delivery, server load, server availability, file type, and network latency whether the second POP provides improved delivery of the content relative to the first POP.
14. A method of balancing load across a CDN having at least a first POP and a second POP, the first POP comprising a first set of edge servers located proximally to a first region, the second POP comprising a second set of edge servers proximally located to a second region that is geographically separated from the first region, the method comprising:
providing a first load balancing server at the first POP for balancing a load at least across the first set of edge servers;
assigning to the first load balancing server, a particular IP address that is also assigned to a second load balancing server, the first load balancing server for balancing a load resulting from requests for content that are resolved to the first POP by Anycast routing as a result of the requests being addressed to the particular IP address and originating from the first region, the second load balancing server for balancing a load resulting from requests for the content that are resolved to the second POP by Anycast routing as a result of the requests being addressed to the particular IP address and originating from the second region;
receiving at the first load balancing server, a request for the content that is addressed to the particular IP address and that originates from the first region;
processing the request at the first load balancing server to determine which edge server of the first and second sets of edge servers optimally delivers the content based on a set of performance parameters;
selecting a particular edge server from the second set of edge servers to satisfy the received request by delivering the content when the particular edge server optimally delivers the content relative to other servers of the first and second sets of servers; and
forwarding the received request from the first load balancing server at the first POP to the selected edge server at the second POP for improved delivery of the content.
15. The method of claim 14 , wherein the selected edge server, in response to receiving the content request, sends the requested content to a requestor that originates the content request.
16. The method of claim 14 , wherein forwarding the content request comprises forwarding the content request from the first load balancing server to the second load balancing server for the second load balancing server to forward the content request to the selected edge server.
17. The method of claim 14 further comprising monitoring at least one of edge server load, edge server availability, file type, network latency, and speed of transfer.
18. The method of claim 17 , wherein said selection of the edge server is based on said monitoring identifying that the selected edge server provides improved delivery for the requested content over a server of the first set of servers.