DATA CENTER NETWORK ARCHITECTURE

• US 20130108263A1
• Filed: 10/12/2012
• Published: 05/02/2013
• Est. Priority Date: 11/01/2011
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Plexxi, Inc. (Hewlett-Packard Enterprise Company)
• Inventors: Srinivas, Anand, deRuijter, Denis H., Husak, David J.
• Granted Patent: US 9,288,555 B2
• Time in Patent Office: Days
• US Class Current: 398/45
• CPC Class Codes: H04B 10/0773 Network aspects, e.g. central monitoring of transmission parameters
• CPC Class Codes: H04L 12/28 characterised by path configuration, e.g. LAN [Local Area Networks] or WAN [Wide Area Networks] wireless communication networks H04W ; arrangements for dividing the transmission path H04W40/00
• CPC Class Codes: H04L 41/0668 by dynamic selection of recovery network elements, e.g. replacement by the most appropriate element after failure
• CPC Class Codes: H04L 41/0816 the condition being an adaptation, e.g. in response to network events
• CPC Class Codes: H04L 41/12 Discovery or management of network topologies
• CPC Class Codes: H04L 69/324 in the data link layer [OSI layer 2], e.g. HDLC
• CPC Class Codes: H04Q 3/0083 Network planning or design; Modelling of planned or existing networks

Abstract

Data center network architectures that can reduce the cost and complexity of data center networks. The data center network architectures can employ optical network topologies and optical nodes to efficiently allocate bandwidth within the data center networks, while reducing the physical interconnectivity requirements of the data center networks. The data center network architectures also allow computing resources within data center networks to be controlled and provisioned based at least in part on a combined network topology and application component topology, thereby enhancing overall application program performance.

Claims

1. A data center network, comprising:

a plurality of nodes interconnected with one another to form the network, the network having a current network state, at least two of the plurality of nodes each having a local controller; and

a central controller communicably coupled to each local controller,

wherein the central controller is operative to provide one or more end-to-end forwarding topologies to each local controller, each end-to-end forwarding topology identifying one or more network segments for forwarding traffic through the network, and

wherein each local controller is independently operative to utilize the end-to-end forwarding topologies to deterministically arrive at a consistent end-to-end forwarding configuration for the network as a function of the current network state.

2. The network of claim 1 further comprising:

at least one other central controller; and

a governing central controller communicably coupled to the central controller and the at least one other central controller, the governing central controller being operative to manage one or more specified operations of the central controller and the at least one other central controller.

3. The network of claim 2 wherein the governing central controller is further operative to manage at least load balancing operations of the central controller and the at least one other central controller.

4. The network of claim 1 further comprising:

a control channel operatively connected between respective local controllers.

5. The network of claim 4 further comprising:

a control plane in which the respective local controllers are communicably interconnected by the control channel.

6. The network of claim 4 wherein the plurality of nodes each have at least one of a packet switch, a circuit switch, and an optical switch for forwarding traffic through the network.

7. The network of claim 6 wherein the respective local controllers are operative to employ the control channel to perform one or more of:

exchanging connectivity information to learn at least a portion of a network topology involving the respective local controllers;

synchronizing the current network state with at least the other respective local controllers using at least the connectivity information;

establishing a logical break on the control channel to filter at least some of the traffic forwarded through the network, thereby avoiding creation of a bridge loop on the control channel;

determining a location of the logical break on the control channel;

propagating one or both of connect notifications and disconnect notifications on the network;

learning at least one endpoint address associated with at least one host computer communicably coupled to the network;

announcing the endpoint address associated with the host computer communicably coupled to the network; and

removing or updating endpoint address entries at the respective local controllers.

8. The network of claim 7 wherein each end-to-end forwarding topology includes a first forwarding topology identifying one or more first network segments for forwarding traffic through the network, and a second forwarding topology identifying one or more second network segments for forwarding traffic through the network.

9. The network of claim 8 wherein the respective local controllers are further operative to compute and maintain the first network segments identified by the first forwarding topology independent of the central controller.

10. The network of claim 9 wherein the respective local controllers are further operative to establish a flooding break on the network to avoid creation of one or more bridge loops within the first forwarding topology.

11. The network of claim 8 wherein the central controller is further operative to compute and provision the second network segments identified by the second forwarding topology.

12. The network of claim 6 further comprising:

a plurality of communication channels forming a data plane and interconnecting nodes on the network.

13. The network of claim 12 wherein the central controller and each local controller are communicably interconnected in the data plane.

14. The network of claim 1 further comprising:

a modeling component having at least one processor operative to execute at least one program out of at least one memory to model an affinity-network topology, the affinity-network topology representing a logical combination of a network topology associated with the network and an affinity topology, the affinity topology specifying policies and attributes that describe communications between a plurality of application components.

15. The network of claim 14 further comprising:

an affinity-network topology database operative to store the affinity-network topology.

16. The network of claim 14 wherein the information pertaining to the network topology includes one or more of circuit switch settings, wavelength settings, a layer-1 link status, a layer-2 link status, endpoint addresses for the network segments, endpoint addresses for one or more host computers communicably coupled to the network, and the current network state.

17. The network of claim 14 wherein the information pertaining to the affinity topology includes information pertaining to one or more of the respective application components, one or more groupings of the application components, one or more links between the respective groupings of the application components, and one or more requirements and attributes associated with the links between the respective groupings of the application components.

18. The network of claim 14 wherein the central controller includes a topology fitting engine operative to receive information pertaining to one or more of the current network state, the affinity topology, one or more network statistics, and one or more configuration capabilities of the respective nodes, and to compute the end-to-end forwarding topologies based at least in part on one or more of the current network state, the affinity topology, the network statistics, and the configuration capabilities of the respective nodes.

19. The network of claim 18 wherein the topology fitting engine is further operative to access information pertaining to at least the affinity topology through a workload abstraction layer interface.

20. The network of claim 19 wherein the topology fitting engine is further operative to provide information pertaining to the end-to-end forwarding topologies through a network abstraction layer interface.

21. The network of claim 20 wherein the information pertaining to the end-to-end forwarding topologies includes one or more of circuit switch settings, a list of candidate forwarding trees provisioned to each node, first weightings assigned to the respective candidate forwarding trees, and mappings of destination addresses to the respective candidate forwarding trees.

22. The network of claim 21 wherein the information pertaining to the end-to-end forwarding topologies includes one or more abstract endpoint definitions, the abstract endpoint definitions including definitions for one or more destination endpoints, and wherein the topology fitting engine is further operative to define the abstract endpoint definitions for the respective nodes, and to map the destination endpoints to the respective candidate forwarding trees based at least in part on the first weightings assigned to the respective candidate forwarding trees.

23. The network of claim 22 wherein each abstract endpoint definition includes one or more of a set of endpoint addresses for one or more host computers communicably coupled to the network, a set of traffic flows, a set of tuples including one or more of uplink/access ports and VLAN IDs for the respective nodes, and a set of uplink/access ports for the respective nodes.

24. The network of claim 20 wherein the information pertaining to the end-to-end forwarding topologies includes one or more of circuit switch settings, a list of candidate forwarding paths provisioned to each node, second weightings assigned to the respective candidate forwarding paths, and mappings of traffic flows to the respective candidate forwarding paths.

25. The network of claim 24 wherein the information pertaining to the end-to-end forwarding topologies includes one or more abstract endpoint definitions, the abstract endpoint definitions including definitions for one or more source/destination endpoint pairs, and wherein the topology fitting engine is further operative to define the abstract endpoint definitions for the respective nodes, and to map the source/destination endpoint pairs to the respective candidate forwarding paths based at least in part on the second weightings assigned to the respective candidate forwarding paths.

26. The network of claim 25 wherein each abstract endpoint definition includes one or more of a set of endpoint addresses for one or more host computers communicably coupled to the network, a set of traffic flows, a set of tuples including one or more of uplink/access ports and VLAN IDs for the respective nodes, and a set of uplink/access ports for the respective nodes.

27. The network of claim 14 wherein the central controller is further operative to provide the end-to-end forwarding topologies to each local controller to pre-provision the end-to-end forwarding topologies at the respective local controllers, each end-to-end forwarding topology being based at least in part on the affinity topology and the current network state, and wherein each local controller is independently operative to deterministically select and utilize at least one of the end-to-end forwarding topologies as an initial forwarding topology, and to maintain the non-selected end-to-end forwarding topologies as alternative forwarding topologies, and, in response to a change in the network state, to deselect the initial forwarding topology at each local controller, and to independently and deterministically select and utilize at least one of the alternative forwarding topologies at each local controller in place of the initial forwarding topology, wherein the at least one of the alternative forwarding topologies selected at each local controller are the same alternative forwarding topologies.

28. The network of claim 1 wherein the network has an associated network topology, the network topology being a physical or logical topology, the network topology being at least one of a linear topology, a ring topology, a chordal ring topology, a multidimensional chordal ring topology, and a 2-dimensional or higher torus topology.

29. The network of claim 1 wherein each node is an optical node.

30. A method of operating a data center network for providing enhanced application program performance and network utilization, comprising the steps of:

providing, by a central controller, one or more end-to-end forwarding topologies to a plurality of local controllers, each local controller being associated with one of a plurality of nodes, the plurality of nodes being interconnected with one another to form the network, each end-to-end forwarding topology identifying one or more network segments for forwarding traffic through the network, the network having a current network state; and

independently utilizing, by the respective local controllers, the end-to-end forwarding topologies to deterministically arrive at a consistent end-to-end forwarding configuration for the network as a function of the current network state.

31. The method of claim 30 further comprising:

managing, by a governing central controller, one or more specified operations of the central controller.

32. The method of claim 30 further comprising:

operatively interconnecting the respective local controllers in a control plane by a control channel.

33. The method of claim 32 wherein the plurality of nodes each have a packet switch, and wherein the method further comprises:

operatively interconnecting the respective packet switches in a data plane by a plurality of communication channels.

34. The method of claim 33 further comprising:

employing, by the respective local controllers, the control channel to perform one or more of;

exchanging connectivity information to learn at least a portion of a network topology involving the respective local controllers;

synchronizing the current network state with at least the other respective local controllers using at least the connectivity information;

establishing a logical break on the control channel to filter at least some of the traffic forwarded through the network, thereby avoiding creation of a bridge loop on the control channel;

determining a location of the logical break on the control channel;

propagating one or both of connect notifications and disconnect notifications on the network;

learning at least one endpoint address associated with at least one host computer communicably coupled to the network;

announcing the endpoint address associated with the host computer communicably coupled to the network; and

removing or updating endpoint address entries at the respective local controllers.

35. The method of claim 34 wherein each end-to-end forwarding topology includes a first forwarding topology identifying one or more first network segments for forwarding traffic through the network, and a second forwarding topology identifying one or more second network segments for forwarding traffic through the network, and wherein the method further comprises:

managing, by at least one of the respective local controllers, the first network segments identified by the first forwarding topology.

36. The method of claim 35 further comprising:

establishing, by at least one of the respective local controllers, a flooding break on the network to avoid creation of one or more bridge loops within the first forwarding topology.

37. The method of claim 36 further comprising:

managing, by the central controller, the second network segments identified by the second forwarding topology.

38. The method of claim 30 further comprising:

modeling an affinity-network topology that represents a logical combination of a network topology associated with the network and an affinity topology, the affinity topology specifying policies and attributes that describe communications between a plurality of application components.

39. The method of claim 38 further comprising:

storing the affinity-network topology in an affinity-network topology database.

40. The method of claim 39 wherein the central controller includes a topology fitting engine, and wherein the method further comprises:

receiving, at the topology fitting engine, the affinity-network topology from the affinity-network topology database; and

computing, by the topology fitting engine, the end-to-end forwarding topologies based at least in part on the affinity-network topology.

41. The method of claim 38 wherein the central controller includes a topology fitting engine, and wherein the method further comprises:

receiving, at the topology fitting engine, information pertaining to one or more of the current network state, the affinity topology, one or more network statistics, and one or more configuration capabilities of the respective nodes; and

computing, by the topology fitting engine, the end-to-end forwarding topologies based at least in part on one or more of the current network state, the affinity topology, the network statistics, and the configuration capabilities of the respective nodes.

42. The method of claim 38 further comprising:

providing, by the central controller, the end-to-end forwarding topologies to each local controller to pre-provision the end-to-end forwarding topologies at the respective local controllers, each end-to-end forwarding topology being based at least in part on the affinity topology and the current network state;

independently and deterministically selecting and utilizing at least one of the end-to-end forwarding topologies as an initial forwarding topology at each local controller;

maintaining the non-selected end-to-end forwarding topologies as alternative forwarding topologies at each local controller;

in response to a change in the network state, deselecting the initial forwarding topology at each local controller; and

independently and deterministically selecting and utilizing at least one of the alternative forwarding topologies at each local controller in place of the initial forwarding topology, wherein the at least one of the alternative forwarding topologies selected at each local controller are the same alternative forwarding topologies.