Resilient hashing with multiple hashes

1. A method of packet forwarding comprising:

maintaining, by a network device, a first hash table comprising mappings between a first set of hash indices and a set of bit values, wherein for each mapping in the first hash table:

the mapping's hash index corresponds to one of a group of next-hop destinations to which network packets received by the network device may be transmitted, and

the mapping's bit value indicates whether the mapping's hash index corresponds to an active or inactive next-hop destination;

maintaining, by the network device, a second hash table comprising mappings between a second set of hash indices and active next-hop destinations in the group of next-hop destinations;

receiving, by the network device, a network packet;

computing, by the network device, a first hash value using a portion of the network packet and a first hash function;

matching, by the network device, the first hash value to a first mapping in the first hash table based on the first mapping's hash index; and

when the first mapping's bit value indicates that the first mapping's hash index corresponds to an active next-hop destination:

matching, by the network device, the first hash value to a second mapping in the second hash table based on the second mapping's hash index; and

sending, by the network device, the network packet to the second mapping's active next-hop destination.

2. The method of claim 1 further comprising, when the first mapping's bit value indicates that the first mapping's hash index corresponds to an inactive next-hop destination:

computing a second hash value using the portion of the network packet and a second hash function that is different from the first hash function;

matching the second hash value to a third mapping in the second hash table based on the third mapping's hash index; and

sending the network packet to the third mapping's active next-hop destination.

3. The method of claim 1 further comprising:

maintaining a third hash table comprising mappings that are identical to the first hash table; and

when the first mapping's bit value indicates that the first mapping's hash index corresponds to an inactive next-hop destination:

computing a second hash value using the portion of the network packet and a second hash function that is different from the first hash function;

matching the second hash value to a third mapping in the third hash table based on the third mapping's hash index; and

when the third mapping's bit value indicates that the third mapping's hash index corresponds to an active next-hop destination:

matching the second hash value to a fourth mapping in the second hash table based on the fourth mapping's hash index; and

sending the network packet to the fourth mapping's active next-hop destination.

4. The method of claim 3 further comprising, when the third mapping's bit value indicates that the third mapping's hash index corresponds to an inactive next-hop destination:

computing a third hash value using the portion of the network packet and a third hash function that is different from the first hash function and the second hash function;

matching the third hash value to a fifth mapping in the second hash table based on the fifth mapping's hash index; and

sending the network packet to the fifth mapping's active next-hop destination.

5. The method of claim 1 wherein a size of the first hash table is equal to a maximum group size for the group of next-hop destinations.

6. The method of claim 5 wherein the maximum group size is defined in a user-configurable file maintained on the network device.

7. The method of claim 6 wherein the user-configurable file further defines:

which next-hop destinations in the group of next-hop destinations are active; and

which next-hop destinations in the group of next-hop destinations are inactive.

8. The method of claim 5 wherein computing the first hash value comprises:

extracting a 5-tuple from the network packet including the network packet's source Internet Protocol (IP) address, source port, destination IP address, destination port, and protocol;

applying the 5-tuple to the first hash function to generate an intermediate hash value; and

computing the intermediate hash value modulo the maximum group size to generate the first hash value.

9. The method of claim 5 wherein a size of the second hash table is equal to the maximum group size multiplied by a replication factor.

10. The method of claim 9 wherein computing the second hash value comprises:

extracting a 5-tuple from the network packet including the network packet's source Internet Protocol (IP) address, source port, destination IP address, destination port, and protocol;

applying the 5-tuple to the second hash function to generate an intermediate hash value; and

computing the intermediate hash value modulo (the maximum group size multiplied by the replication factor) to generate the second hash value.

11. A network device comprising:

a processor configured to:

maintain a first hash table comprising mappings between a first set of hash indices and a set of bit values, wherein for each mapping in the first hash table:

the mapping's hash index corresponds to one of a group of next-hop destinations to which network packets received by the network device may be transmitted, and

the mapping's bit value indicates whether the mapping's hash index corresponds to an active or inactive next-hop destination;

maintain a second hash table comprising mappings between a second set of hash indices and active next-hop destinations in the group of next-hop destinations;

receive a network packet;

compute a first hash value using a portion of the network packet and a first hash function;

match the first hash value to a first mapping in the first hash table based on the first mapping's hash index; and

when the first mapping's bit value indicates that the first mapping's hash index corresponds to an active next-hop destination:

match the first hash value to a second mapping in the second hash table based on the second mapping's hash index; and

send the network packet to the second mapping's active next-hop destination.

12. The network device of claim 11 wherein when the first mapping's bit value indicates that the first mapping's hash index corresponds to an inactive next-hop destination, the processor is further configured to:

compute a second hash value using the portion of the network packet and a second hash function that is different from the first hash function;

match the second hash value to a third mapping in the second hash table based on the third mapping's hash index; and

send the network packet to the third mapping's active next-hop destination.

13. The network device of claim 11 wherein the processor comprises an application-specific integrated circuit (ASIC).

14. The network device of claim 11 wherein the processor comprises a general purpose central processing unit (CPU).

15. The network device of claim 13 wherein the first hash table is implemented as a bit vector stored on the ASIC.

16. The network device of claim 13 wherein the second hash table is implemented as a direct index table stored on the ASIC.

17. A non-transitory computer readable storage medium having stored thereon program code executable by a network device, the program code comprising:

code that causes the network device to maintain a first hash table comprising mappings between a first set of hash indices and a set of bit values, wherein for each mapping in the first hash table:

the mapping's hash index corresponds to one of a group of next-hop destinations to which network packets received by the network device may be transmitted, and

the mapping's bit value indicates whether the mapping's hash index corresponds to an active or inactive next-hop destination;

code that causes the network device to maintain a second hash table comprising mappings between a second set of hash indices and active next-hop destinations in the group of next-hop destinations;

code that causes the network device to receive a network packet;

code that causes the network device to compute a first hash value using a portion of the network packet and a first hash function;

code that causes the network device to match the first hash value to a first mapping in the first hash table based on the first mapping's hash index; and

when the first mapping's bit value indicates that the first mapping's hash index corresponds to an active next-hop destination:

code that causes the network device to match the first hash value to a second mapping in the second hash table based on the second mapping's hash index; and

code that causes the network device to send the network packet to the second mapping's active next-hop destination.

18. The non-transitory computer readable storage medium of claim 17 wherein the group of next-hop destinations correspond to a plurality of servers associated with a service,

wherein the network device is a network load balancer configured to load balance, among the plurality of servers, network traffic destined for the service; and

wherein the program code further comprises:

code that causes the network device to maintain a version table comprising mappings between version keys and version identifiers; and

code that causes the network device to maintain a history table comprising mappings between history keys and active next-hop destinations in the group of next-hop destinations.

19. The non-transitory computer readable storage medium of claim 18 wherein the program code further comprises, prior to transmitting the network packet to the third mapping's active next-hop destination:

code that causes the network device to combine the first hash value and the second hash value into a version key for the network packet; and

code that causes the network device to match the version key for the network packet to a fourth mapping in the version table based on the fourth mapping's version key, and

wherein the code that causes the network device to send the network packet to the third mapping's active next-hop destination comprises:

code that causes the network device to encapsulate the network packet with a header that includes the fourth mapping's version identifier; and

code that causes the network device to transmit the encapsulated network packet to the third mapping's active next-hop destination.

20. The non-transitory computer readable storage medium of claim 19 wherein the program code further comprises:

code that causes the network device to receive the encapsulated network packet back from the third mapping's active next-hop destination;

code that causes the network device to de-capsulate the network packet from the encapsulated network packet, the de-capsulating including extracting the fourth mapping's version identifier from the header;

code that causes the network device to compute a third hash value based on a portion of the de-capsulated network packet and the first hash function;

code that causes the network device to compute a fourth hash value based on the portion of the de-capsulated network packet and the second hash function;

code that causes the network device to combine the third hash value, the fourth hash value, and the extracted version identifier into a history key for the de-capsulated network packet; and

when the history key for the de-capsulated network packet matches a fifth mapping in the history table based on the fifth mapping's history key:

code that causes the network device to modify the extracted version identifier for future processing;

code that causes the network device to re-encapsulate the de-capsulated network packet with another header that includes the modified version identifier; and

code that causes the network device to transmit the re-encapsulated network packet to the fifth mapping's active next-hop destination.