Network-aware load balancing
Some embodiments of the invention provide a method for network-aware load balancing for data messages traversing a software-defined wide area network (SD-WAN) (e.g., a virtual network) including multiple connection links between different elements of the SD-WAN. The method includes receiving, at a load balancer in a multi-machine site, link state data relating to a set of SD-WAN datapaths including connection links of the multiple connection links. The load balancer, in some embodiments, provides load balancing for data messages sent from a machine in the multi-machine site to a set of destination machines (e.g., web servers, database servers, etc.) connected to the load balancer over the set of SD-WAN datapaths. The load balancer selects, for the data message, a particular destination machine (e.g., a frontend machine for a set of backend servers) in the set of destination machines by performing a load balancing operation based on the received link state data.
1. A method of providing network-aware load balancing for data messages traversing a software-defined wide area network (SD-WAN) comprising a plurality of connection links between different elements of the SD-WAN, the method comprising:
at a load balancer,
receiving link state data relating to a set of SD-WAN datapaths comprising a set of connection links in the plurality of connection links, wherein the load balancer provides load balancing for a set of destination machines connected to the load balancer over the set of SD-WAN datapaths;
receiving a data message from a device connected to the SD-WAN destined to a destination machine in the set of destination machines;
selecting, for the data message, a particular destination machine in the set of destination machines by performing a load balancing operation based on the received link state data, the link state data comprising attributes of different types of physical links; and
forwarding the data message to the selected particular destination machine in the set of destination machines.
2. The method of claim 1 , wherein the load balancer receives the link state data from a controller of the SD-WAN.
3. The method of claim 1 , wherein the load balancer receives the link state data from a collocated SD-WAN device.
4. The method of claim 1 further comprising receiving load information regarding each destination machine in the set of destination machines, wherein the load balancing operation is further based on the received load information.
5. The method of claim 4 , wherein the load balancing operation comprises computing a weight associated with each destination machine in the set of destination machines based on (i) the link state data and (ii) the received load information.
6. The method of claim 4 , wherein the load balancing operation comprises:
a first load balancing sub-operation based on the received load information that selects a set of candidate destination machines in the set of destination machines; and
a second load balancing sub-operation based on the link state data to select the particular destination machine from the set of candidate destination machines.
7. The method of claim 6 , wherein the second load balancing sub-operation based on the link state data further selects a particular datapath associated with a candidate destination machine.
8. The method of claim 1 , wherein the set of destination machines is a set of SD-WAN edge devices that each provide access to resources associated with the SD-WAN edge device.
9. The method of claim 8 , wherein each SD-WAN edge device in the set of SD-WAN edge devices is associated with a data center in which the resources associated with the SD-WAN edge device execute.
10. The method of claim 1 , wherein the set of destination machines comprises a set of frontend load balancers for a set of backend compute nodes.
11. The method of claim 1 , wherein the set of datapaths comprises a subset of multiple datapaths to a same destination machine in the set of destination machines, wherein at least two datapaths in the subset of multiple datapaths traverse at least one of different SD-WAN hubs, and different SD-WAN gateways.
12. The method of claim 1 , wherein the set of multiple datapaths to the same destination machine comprise at least two datapaths that traverse a same set of components of the SD-WAN in different orders.
13. The method of claim 1 , wherein a plurality of the connection links that are included in the set of SD-WAN datapaths are tunnels between two elements of the SD-WAN and the link state data is based on characteristics of the tunnel and the elements of the SD-WAN that terminate the tunnels.
14. The method of claim 1 , wherein the received link state data is a set of weights used to perform the load balancing operation.
15. The method of claim 14 , wherein the set of weights comprises, for each destination machine, (1) a first load weight indicating at least one of a CPU load, a memory load, and a session load on the destination machine and (2) a second network weight associated with a set of datapaths connecting the load balancer to the destination machine.
16. The method of claim 14 , wherein the set of weights comprises, for each datapath in the set of datapaths, (1) a first load weight indicating at least one of a CPU load, a memory load, and a session load on the associated destination machine and (2) a second network weight associated with a particular set of connection links making up the datapath based on the received connection link characteristic data.
17. The method of claim 1 , wherein the link state data comprises at least one of a current measure of latency for each datapath in the plurality of datapaths, a current measure of latency for each datapath in the plurality of datapaths, a current measure of jitter for each datapath in the plurality of datapaths, and a current measure of quality of experience score for each datapath in the plurality of datapaths.
18. The method of claim 17 , wherein the link state data further comprises at least one of a historical measure of latency for each datapath in the plurality of datapaths, a historical measure of latency for each datapath in the plurality of datapaths, a historical measure of jitter for each datapath in the plurality of datapaths, and a historical measure of quality of experience score for each datapath in the plurality of datapaths.
19. The method of claim 1 , wherein the link state data comprises at least one of a current measure of latency for each connection link in the set of connection links, a current measure of latency for each connection link in the set of connection links, a current measure of jitter for each connection link in the set of connection links, and a current measure of quality of experience score for each connection link in the set of connection links.
20. The method of claim 19 , wherein the link state data further comprises at least one of a historical measure of latency for each connection link in the set of connection links, a historical measure of latency for each connection link in the set of connection links, a historical measure of jitter for each connection link in the set of connection links, and a historical measure of quality of experience score for each connection link in the set of connection links.
21. A non-transitory machine readable medium storing a program for execution by at least one processor, the program for performing any one of the method claims 1-20 .