Systems and methods for traffic management using load metrics reflective of a requested service
A method for determining metrics of a content delivery and global traffic management network provides service metric probes that determine the service availability and metric measurements of types of services provided by a content delivery machine. Latency probes are also provided for determining the latency of various servers within a network. Service metric probes consult a configuration file containing each DNS name in its area and the set of services. Each server in the network has a metric test associated with each service supported by the server which the service metric probes periodically performs metric tests on and records the metric test results which are periodically sent to all of the DNS servers in the network. DNS servers use the test result updates to determine the best server to return for a given DNS name. The latency probe calculates the latency from its location to a client's location using the round trip time for sending a packet to the client to obtain the latency value for that client. The latency probe updates the DNS servers with the clients' latency data. The DNS server uses the latency test data updates to determine the closest server to a client.
1. A system, comprising:
a first machine managed by a service provider and comprising at least one hardware microprocessor coupled to memory storing computer program instructions that are executable on the at least one hardware microprocessor to provide a particular service that sends requested content to requesting clients via a network interface;
a second machine managed by the service provider and comprising at least one hardware microprocessor coupled to memory storing computer program instructions that are executable on the at least one hardware microprocessor to provide a DNS service via a network interface, wherein upon execution of the computer program instructions the DNS service:
receives a request to resolve a hostname from a client;
selects a particular load metric from among a plurality of load metrics available to the DNS service, the selection being based at least in part on the hostname, wherein the particular load metric reflects load on the particular service;
determines an IP address to return to the client in response to the client's request to resolve the hostname, wherein said determination is based at least in part on the particular load metric.
2. The system of claim 1 , further comprising a custom service probe that measures the particular load metric on each of a plurality of machines providing the particular service, the custom service probe being implemented with computer program instructions that are executable by the first machine or by another machine managed by the service provider.
3. The system of claim 1 , wherein the particular load metric is specific to the particular service.
4. The system of claim 1 , wherein the particular load metric is based at least in part on a time for the first machine to process a service request for the particular service.
5. The system of claim 1 , wherein the particular load metric is based at least in part on a time to retrieve an html file for a web page using the particular service.
6. The system of claim 1 , wherein the first machine comprises memory storing computer program instructions that are executable on the at least one hardware microprocessor to run the machine as a cache server storing content for delivery to requesting clients.
7. The system of claim 1 , wherein the first machine comprises memory storing computer program instructions that are executable on the at least one hardware microprocessor to run the machine as a web server.
8. The system of claim 1 , wherein the client from which the request to resolve the hostname was received is a machine running as a local DNS server.
9. A method, comprising:
receiving requests for content from clients, the requests being received by a first machine managed by a service provider and comprising at least one hardware microprocessor coupled to memory storing computer program instructions that are executable on the at least one hardware microprocessor to provide a particular service that sends requested content to requesting clients via a network interface;
with a second machine managed by the service provider and comprising at least one hardware microprocessor coupled to memory storing computer program instructions that are executable on the at least one hardware microprocessor to provide a DNS service via a network interface:
receiving a request to resolve a hostname from a client;
selecting a particular load metric from among a plurality of load metrics, the selection being based at least in part on the hostname, wherein the particular load metric reflects load on the particular service;
determining an IP address to return to the client in response to the client's request to resolve the hostname, wherein said determination is based at least in part on the particular load metric.
10. The method of claim 9 , wherein the particular load metric is specific to the particular service.
11. The method of claim 9 , wherein the particular load metric is based at least in part on a time for the first machine to process a service request for the particular service.
12. The method of claim 9 , wherein the particular load metric is based at least in part on a time to retrieve an html file for a web page using the particular service.
13. The method of claim 9 , wherein the first machine is a machine running as a cache server.
14. The method of claim 9 , wherein the first machine is a machine running as a web server.
15. The method of claim 9 , wherein the client from which the request to resolve the hostname was received is a machine running as a local DNS server.