Load management for servers in shared address network architecture
A system for load management in a shared address networking architecture includes a primary point-of-presence (POP) group of servers configured to serve content of a domain and that are each reachable at an address of a first IP address block and a secondary PoP group of servers configured to serve the content of the domain and that are each reachable at an address of a second IP address block. The system further includes a traffic management agent configured to reduce a total volume of incoming requests received by the primary PoP group for a period of time following a return of a first server in the primary PoP group of servers to an online state by selectively directing a first percentage of the incoming requests to the second IP address block instead of the first IP address block.
1. A system comprising:
a primary point-of-presence (POP) group of servers including multiple server pools in different geographic locations, each server in the primary POP group being reachable at an address of a first internet protocol (IP) address block and configured to serve content of a domain;
a secondary PoP group of servers, each server in the secondary POP group being reachable at an address of a second IP address block and configured to serve the content of the domain; and
a traffic management agent stored in memory and configured to reduce a total volume of incoming requests received by the primary POP group for a period of time following a return of a first POP server pool in the primary PoP group of servers to an online state by selectively directing a first percentage of the incoming requests to the second IP address block instead of the first IP address block, and wherein the traffic management agent gradually decreases the first percentage of incoming requests directed to the second IP address block as time passes following the return of the first POP server to the online state.
2. The system of claim 1 , further comprising:
a health evaluator stored in the memory and configured to compute a health metric for the primary POP group, wherein the traffic management agent determines the first percentage of requests to direct away from the primary POP group of servers based on the health metric.
3. The system of claim 2 , wherein the health metric for the primary PoP group of servers is computed based on health data received from the first POP server pool.
4. The system of claim 3 , wherein the health data includes central processing unit (CPU) usage and memory usage for the first POP server pool.
5. The system of claim 2 , wherein the traffic management agent and the health evaluator are both executed by a same DNS server.
6. The system of claim 1 , wherein the secondary PoP group of servers includes fewer servers than the primary PoP group of servers and wherein the servers of the secondary PoP group are less geographically dispersed than the servers of the primary PoP group.
7. The system of claim 6 , and wherein the traffic management agent directs substantially all incoming requests for the domain to the secondary POP group of servers for a first time interval immediately following the return of the first POP server pool to the online state.
8. The system of claim 1 , wherein the traffic management agent directs substantially all of the incoming requests to the primary POP group of servers after a threshold period of time has elapsed following the return of the first POP server pool to the online state.
9. A method comprising:
receiving, at a DNS server, requests to access a domain associated with both:
a primary point-of-presence (POP) group of servers for the domain, the primary PoP group of servers including multiple server pools in different geographic locations, each of the servers in the primary POP group of servers being reachable via an address of a first internet protocol (IP) address block; and
a secondary PoP group of servers for the domain, each server in the secondary PoP group being reachable via an address of a second IP address block;
detecting return of a first POP server pool in the primary POP to an online state after the first POP server pool has been offline for a period of time; and
reducing a total volume of incoming requests received by the first POP group of servers by selectively directing, by the DNS server, a first percentage of the incoming requests to the second IP address block instead of the first IP address block for a period of time following the return of the first POP server pool to the online state, wherein the first percentage of incoming requests directed to the second IP address block is gradually decreased as time passes following the return of the first POP server to the online state.
10. The method of claim 9 , further comprising:
computing a health metric for the primary POP group of servers; and
determining the first percentage based on the health metric.
11. The method of claim 10 , wherein the health metric for the primary PoP group is computed based on health data received from the first POP server pool.
12. The method of claim 11 , wherein the health data includes central processing unit (CPU) usage and memory usage for the first POP server pool.
13. The method of claim 11 , wherein substantially all of the incoming requests for the domain are directed to the primary POP group after a threshold period of time has elapsed following the return of the first POP server pool to the online state.
14. The method of claim 9 , wherein the secondary PoP group includes fewer servers than the primary PoP group and wherein the secondary POP group is less geographically dispersed than the primary POP group.
15. The method of claim 9 , further comprising directing substantially all incoming requests for the domain to the secondary POP group for a first time interval immediately following the return of the first POP server pool to the online state.
16. One or more computer-readable storage media encoding computer-executable instructions for executing a computer process, the computer process comprising:
determining that a first point-of-presence (PoP) server pool within a primary point-of-presence (POP) group of servers has returned to an online state within a defined recency time interval, the primary PoP group including multiple server pools in different geographic locations, each server of the multiple server pools sharing a first internet protocol (IP) address block and being configured to serve content of a domain;
determining a first percentage of incoming requests for the domain to direct to a second IP address block instead of the first IP address block in response to the return of the POP server pool to the online state, the second IP address block being shared by a secondary POP group of servers configured to serve the content of the domain; and
directing the first percentage of incoming requests to the second IP address block instead of the first IP address block for a threshold period of time following the return of the first POP server pool to the online state; and
gradually decreasing the first percentage of the incoming requests directed to the second IP address block while gradually increasing a second percentage of the incoming requests directed to the first IP address block as time passes following the return of the first POP server pool to the online state.
17. The one or more computer-readable storage media of claim 16 , wherein the computer process further comprises:
determining a health metric characterizing the primary PoP group of servers; and
determining the first percentage of requests to direct away from the primary POP group of servers based on the health metric.