Adaptive rate limiting of flow probes
A technique for performing adaptive rate limiting of flow probes is described. The technique includes sending a plurality of flow probes from a first service engine to at least one other service engine. The flow probes are sent at a rate that does not exceed a rate limit. A flow probe of the plurality of flow probes is generated in response to the first service engine receiving a mid-flow packet for a flow that is not recognized by the first service engine. A recipient service engine of the flow probe responds with a success indicator if the recipient service engine recognizes the flow. The technique also includes determining a success rate associated with success indicators received from the at least one other service engine and comparing the success rate with a first threshold. The rate limit is adjusted in response to a comparison result.
1. A method of detecting an attack on a network comprising a plurality of service engine, the method comprising:
at a first service engine:
detecting a first plurality of flows as flows that the first service engine has not seen before;
based on a flow-probe first rate, sending flow probes to a set of other service engines to determine whether the set of other service engines has seen the first plurality of flows;
based on a first success rate that is associated with success indicators received from the set of service engine and that is less than a threshold rate, reducing a rate of flow probes from the first rate to a lower, second rate in order to account for a possibility that the network is under attack.
2. The method of claim 1 further comprising increasing the flow-probe rate from the second rate to a higher, third rate when for a second plurality of flows, the set of service engines returns success indicators at a second success rate higher than the first success rate.
3. The method of claim 1 , wherein reducing the rate comprises reducing the rate when the success rate is less than the threshold rate for a duration of time.
4. The method of claim 1 further comprising mapping each flow in the plurality of flows to one service engine to receive a flow probe for the flow.
5. The method of claim 1 , wherein detecting each flow in the first plurality of flows comprises detecting a packet for each flow that is a mid-flow packet for the flow that has not been previously received by the first service engine.
6. The method of claim 5 , wherein mid-flow packets are packets sent by one or more attackers to the network, and responses generated for such packets are intended to strain resources of the network.
7. The method of claim 1 , wherein the plurality of service engines are between clients and backend servers and perform service operations on requests sent by the clients to the backend servers.
8. The method of claim 7 , wherein the plurality of service engines are load balancing service engines.
9. The method of claim 7 , wherein the plurality of service engines execute on host computers along with backend servers.
10. The method of claim 1 , wherein the plurality of service engines are configured with the flow-probe rates and threshold value by a controller operating in the network.
11. A non-transitory machine readable medium storing a program for detecting an attack on a network comprising a plurality of service engine, the program executable by a set of processing units, the program comprising sets of instructions for:
at a first service engine:
detecting a first plurality of flows as flows that the first service engine has not seen before;
based on a flow-probe first rate, sending flow probes to a set of other service engines to determine whether the set of other service engines has seen the first plurality of flows;
based on a first success rate that is associated with success indicators received from the set of service engine and that is less than a threshold rate, reducing a rate of flow probes from the first rate to a lower, second rate in order to account for a possibility that the network is under attack.
12. The non-transitory machine readable medium of claim 11 , the program further comprising a set of instructions for increasing the flow-probe rate from the second rate to a higher, third rate when for a second plurality of flows, the set of service engines returns success indicators at a second success rate higher than the first success rate.
13. The non-transitory machine readable medium of claim 11 , wherein the set of instructions for reducing the rate comprises a set of instructions for reducing the rate when the success rate is less than the threshold rate for a duration of time.
14. The non-transitory machine readable medium of claim 11 , the program further comprising a set of instructions for mapping each flow in the plurality of flows to one service engine to receive a flow probe for the flow.
15. The non-transitory machine readable medium of claim 11 , wherein the set of instructions for detecting each flow in the first plurality of flows comprises a set of instructions for detecting a packet for each flow that is a mid-flow packet for the flow that has not been previously received by the first service engine.
16. The non-transitory machine readable medium of claim 15 , wherein mid-flow packets are packets sent by one or more attackers to the network, and responses generated for such packets are intended to strain resources of the network.
17. The non-transitory machine readable medium of claim 11 , wherein the plurality of service engines are between clients and backend servers and perform service operations on requests sent by the clients to the backend servers.
18. The non-transitory machine readable medium of claim 17 , wherein the plurality of service engines are load balancing service engines.
19. The non-transitory machine readable medium of claim 17 , wherein the plurality of service engines execute on host computers along with backend servers.
20. The non-transitory machine readable medium of claim 11 , wherein the plurality of service engines are configured with the flow-probe rates and threshold value by a controller operating in the network.