Method for maintaining reservation state in a network router and resulting scalable integrated architectures for computer networks
View Patent ↗Methods are described for maintaining a small bounded aggregate state within network routers pertaining to selected architectural families. Small bounded reservations states are utilized wherein the method is fully scalable for use on large networks. The size of the aggregate state and the complexity of the associated refresh mechanism is determined by the parameters of the network, such as size, and classes, which is in contrast to states based on the number of end-user flows. The method can render deterministic bandwidth use within the network wherein real-time multimedia applications may be accorded strict delay and bandwidth guarantees. The invention provides a middle-ground between the stateful Intserv and the stateless SCORE architectures.
1. A method for maintaining the reservation state in a network router, comprising maintaining a bounded aggregate per-destination reservation state instead of a per-flow reservation state, wherein the source of a flow sends a refresh message to said router; and wherein all refresh messages of a particular destination are aggregated at said router, and wherein a refresh message specifies a destination and bandwidth for that destination.
2. A method as recited in claim 1 :
wherein said router maintains rates of incoming and outgoing traffic; and
wherein said router does not maintain information on rates of flow.
3. A method as recited in claim 1 , further comprising maintaining a set of token-buckets arranged in the form of a tree for aggregating network flows into classes.
4. A method as recited in claim 1 , further comprising aggregating network flows utilizing burst-drain-time or burst-ratio.
5. A method as recited in claim 1 , further comprising merging a set of data flows into a smaller set of aggregated flows.
6. A method as recited in claim 5 , where in said data flows are merged based on class or destination.
7. A method as recited in claim 1 , wherein said router maintain state only for aggregated flows and processes only aggregated flows.
8. A method as recited in claim 7 , further comprising:
providing guarantees to aggregated flows; and
providing guarantees to individual flows within the aggregated flows.
9. A method as recited in claim 1 , further comprising using diffusing computations to maintain consistency of the reservation.
10. A method as recited in claim 1 , wherein said aggregate state has a size and associated refresh mechanism.
11. A method as recited in claim 10 , wherein aggregate state size and refresh mechanism complexity are a function of a network parameter rather than a function of the number of end-user flows.
12. A method as recited in claim 11 , wherein said network parameter comprises class.
13. A method as recited in claim 11 , wherein said network parameter comprises destination.
14. A method as recited in claim 1 , wherein said step of maintaining aggregate per-destination reservation state instead of a per-flow reservation state comprises storing and refreshing resource reservations on a per-destination basis rather than on a per-flow basis.
15. A method as recited in claim 1 , further comprising utilizing per-destination refresh messages instead of per-flow refresh messages.
16. A method as recited in claim 15 , wherein a refresh message specifies a destination and bandwidth for that destination.
17. A method as recited in claim 16 , wherein when a flow terminates, the source stops sending refresh messages and bandwidth reserved for the flow is released.
18. A method as recited in claim 1 , wherein when a flow terminates, the source stops sending refresh messages and bandwidth reserved for the flow is released.
19. A method for maintaining the reservation state in a network router, comprising storing and refreshing resource reservations on a per-destination basis, rather than on a per-flow basis, wherein the source of a flow sends a refresh message to said router; and wherein all refresh messages of a particular destination are aggregated at said router; and wherein a refresh message specifies a destination and bandwidth for that destination.
20. A method as recited in claim 19 :
wherein said router maintains rates of incoming and outgoing traffic; and
wherein said router does not maintain information on rates of flow.
21. A method as recited in claim 19 , further comprising maintaining a set of token-buckets arranged in the form of a tree for aggregating network flows into classes.
22. A method as recited in claim 19 , further comprising aggregating network flows utilizing burst-drain-time or burst-ratio.
23. A method as recited in claim 19 , further comprising merging a set of data flows into a smaller set of aggregated flows.
24. A method as recited in claim 23 , wherein said data flows are merged based on class or destination.
25. A method as recited in claim 19 , wherein said router maintain state only for aggregated flows and processes only aggregated flows.
26. A method as recited in claim 25 , further comprising:
providing guarantees to aggregated flows; and
providing guarantees to individual flows within the aggregated flows.
27. A method as recited in claim 19 , further comprising using diffusing computations to maintain consistency of the reservation.
28. A method as recited in claim 19 , wherein said aggregate state has a size and associated refresh mechanism.
29. A method as recited in claim 28 , wherein aggregate state size and refresh mechanism complexity are a function of a network parameter rather than a function of the number of end-user flows.
30. A method as recited in claim 29 , wherein said network parameter comprises class.
31. A method as recited in claim 29 , wherein said network parameter comprises destination.
32. A method as recited in claim 19 , further comprising utilizing per-destination refresh messages instead of per-flow refresh messages.
33. A method as recited in claim 32 , wherein a refresh message specifies a destination and bandwidth for that destination.
34. A method as recited in claim 33 , wherein when a flow terminates, the source stops sending refresh messages and bandwidth reserved for the flow is released.
35. A method as recited in claim 19 , wherein when a flow terminates, the source stops sending refresh messages and bandwidth reserved for the flow is released.
36. A method for maintaining the reservation state in a network router, comprising:
merging a set of data flows into a smaller set of aggregated flows; and
storing and refreshing resource reservations on a per-destination basis, rather than on a per-flow basis;
wherein said router maintains rates of incoming and outgoing traffic; and
wherein said router does not maintain information on rates of flow, wherein the source of a flow sends a refresh message to said router; and wherein all refresh messages of a particular destination are aggregated at said router, and wherein a refresh message specifies a destination and bandwidth for that destination.
37. A method as recited in claim 36 , further comprising maintaining a set of token-buckets arranged in the form of a tree for aggregating network flows into classes.
38. A method as recited in claim 40 , further comprising aggregating network flows utilizing burst-drain-time or burst-ratio.
39. A method as recited in claim 36 , wherein said data flows are merged based on class or destination.
40. A method as recited in claim 36 , wherein said router maintains state only for aggregated flows and processes only aggregated flows.
41. A method as recited in claim 40 , further comprising:
providing guarantees to aggregated flows; and
providing guarantees to individual flows within the aggregated flows.
42. A method as recited in claim 36 , further comprising using diffusing computations to maintain consistency of the reservation.
43. A method as recited in claim 36 , wherein said aggregate state has a size and associated refresh mechanism.
44. A method as recited in claim 43 , wherein aggregate state size and refresh mechanism complexity are a function of a network parameter rather than a function of the number of end-user flows.
45. A method as recited in claim 44 , wherein said network parameter comprises class.
46. A method as recited in claim 44 , wherein said network parameter comprises destination.
47. A method as recited in claim 36 , further comprising utilizing per-destination refresh messages instead of per-flow refresh messages.
48. A method as recited in claim 47 , wherein a refresh message specifies a destination and bandwidth for that destination.
49. A method as recited in claim 48 , wherein when a flow terminates, the source stops sending refresh messages and bandwidth reserved for the flow is released.
50. A method as recited in claim 36 , wherein when a flow terminates, the source stops sending refresh messages and bandwidth reserved for the flow is released.