Scalable Commodity Data Center Network Architecture

• US 20100020806A1
• Filed: 07/22/2008
• Published: 01/28/2010
• Est. Priority Date: 07/22/2008
• Status: Active Grant

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Vahdat, Amin, Al-Fares, Mohammad, Loukissas, Alexander
• Granted Patent: US 8,483,096 B2
• Time in Patent Office: Days
• US Class Current: 370/395.310
• CPC Class Codes: H04L 45/00 Routing or path finding of packets in data switching networks routing or path finding in wireless networks H04W40/00
• CPC Class Codes: H04L 45/7453 using hashing
• CPC Class Codes: H04L 49/3009 Header conversion, routing tables or routing tags
• CPC Class Codes: H04L 49/552 by ensuring the integrity of packets received through redundant connections

Abstract

In one aspect, there is provided a method. The method may include receiving a packet at a switch coupled to a plurality of switches. The switch may determine, from a first level table comprising a plurality of prefixes, a prefix matching a first portion of a destination address of the received packet. The switch may also determine, from a second level table comprising a plurality of suffixes, a suffix matching a second portion of a destination address of the received packet, when the matching prefix of the first level table is associated with the second level table. The switch may forward, based on the first level table and the second level table, the received packet to an output port of the switch. In some implementations, the plurality of switches are configured as a fat-tree network. Moreover, in some implementations, the subject matter described herein provides a data center communication fabric that provides scalable communication bandwidth with significant fault tolerance using a plurality of smaller individual switches of substantially the same size and capability. Related systems, apparatus, methods, and/or articles are also described.

Claims

1. A method comprising:

receiving a packet at a switch, the switch coupled to a plurality of switches;

determining, from a first level table comprising a plurality of prefixes, a prefix matching a first portion of a destination address of the received packet;

determining, from a second level table comprising a plurality of suffixes, a suffix matching a second portion of a destination address of the received packet, when the matching prefix of the first level table is associated with the second level table; and

forwarding, based on the first level table and the second level table, the received packet to an output port of the switch.

2. The method of claim 1, wherein receiving further comprises:

configuring the plurality of switches as a fat-tree network.

3. The method of claim 1, wherein determining, from the first level table, further comprises:

determining the prefix by determining a longest matching prefix from the plurality of prefixes of the first level table, the prefix representing one or more most significant bytes of the destination address.

4. The method of claim 1, wherein forwarding further comprises:

forwarding, by the switch using the first level table and the second table, the received packet to a first output port of the switch, the switch implemented as an edge switch, the first output port coupled to an aggregation switch and mapped to the matching prefix; and

forwarding, by the aggregation switch using the first level table and the second table, the received packet to a second output port at the aggregation switch, the second output port coupled to a core switch and mapped to the the matching suffix.

5. The method of claim 1 further comprising:

generating the first level table to include one or more prefixes, the prefixes representing a route.

6. The method of claim 1 further comprising:

generating the second level table to include one or more suffixes, the suffixes representing a route.

7. The method of claim 1, further comprising:

generating the second level table to include one or more suffixes, each of the suffixes representing a host identifier.

8. The method of claim 1 further comprising:

configuring the plurality of switches of the pod to include one or more edge switches coupled to one or more aggregation switches, the aggregation switches of the pod further coupled to one or more core switches, wherein the edge switches, the aggregation switches, and the core switches are configured as a fat-tree network.

9. The method of claim 8 further comprising:

implementing the one or more edge switches, the one or more aggregation switches, and the one or more core switches using a substantially similar configuration with respect to a quantity of ports per switch and a capacity of each of the ports.

10. The method of claim 8 further comprising:

implementing at least one of the edge switches, at least one of the aggregation switches, and at least one of the core switches using the same configuration with respect to a quantity of ports per switch and a capacity of each of the ports.

11. The method of claim 1 further comprising:

forwarding, to one of the output ports of the switch, one or more packets associated with a flow to maintain a sequence of the packets;

scheduling one or more packets associated with a flow; and

monitoring the plurality of switches for a fault.

12. A system comprising:

a switch configured to receive one or more packets, the switch coupled to a plurality of switches, the plurality of switches configured as a pod of a fat-tree network, the switch further configured to forward the one or more packets based on a first level table and a second level table.

13. The system of claim 12, wherein the first level table further comprises:

one or more prefixes representing a route,.

14. The system of claim 12, wherein the second level table further comprises:

one or more suffixes, the suffixes representing a route.

15. The system of claim 12, wherein the second level table further comprises:

one or more suffixes, each of the suffixes representing a host identifier.

16. A network comprising:

a plurality of switches configured as a fat-tree network, the fat-tree network providing communications to a cluster of computers at a data center, each of the plurality of switches including a substantially similar configuration with respect to a quantity of ports per switch and a capacity of each of the ports.

17. The network of claim 16 further comprising:

configuring the plurality of switches to include one or more edge switches coupled to one or more aggregation switches, the aggregation switches further coupled to one or more core switches, wherein the edge switches, the aggregation switches, and core switches are configured as the fat-tree network.

18. The network of claim 17 further comprising:

implementing the one or more edge switches, the one or more aggregation switches, and the one or more core switches using the same configuration with respect to a quantity of ports per switch and a capacity of each of the ports.

19. A system comprising:

a plurality of switches configured as a network providing a plurality of paths to a cluster of computers at a data center, the plurality of switches configured to deliver, from a source to a destination, a plurality of packets in order on a per flow basis, and the plurality of switches configured to distribute flows among the plurality of paths.

20. The system of claim 19 further comprising:

using a second level table to distribute flows among the plurality of paths, the second level table comprising one or more suffixes representing a host identifier.