Low-latency processing of multicast packets

Patent metadata

Field	Value
Publication number	US-11218415-B2
Application number	US-201816194345-A
Country	US
Kind code	B2
Filing date	Nov 18, 2018
Priority date	Nov 18, 2018
Publication date	Jan 4, 2022
Grant date	Jan 4, 2022

Abstract

A network element includes multiple ports and forwarding circuitry. The ports are configured to serve as network interfaces for exchanging packets with a communication network. The forwarding circuitry is configured to receive a multicast packet that is to be forwarded via a plurality of the ports over a plurality of paths through the communication network to a plurality of destinations, to identify a path having a highest latency among the multiple paths over which the multicast packet is to be forwarded, to forward the multicast packet to one or more of the paths other than the identified path, using a normal scheduling process having a first forwarding latency, and to forward the multicast packet to at least the identified path, using an accelerated scheduling process having a second forwarding latency, smaller than the first forwarding latency.

First claim

The invention claimed is:

1. A packet communication network, comprising multiple packet switches connected together in a multi-layer topology, which includes at least a leaf layer, containing at least first and second leaf switches, and a spine layer, containing at least one spine switch, through which the first and second leaf switches are interconnected,
wherein the first leaf switch comprises:
multiple ports, which are configured to serve as network interfaces for exchanging packets with the network and include at least an up-port leading to the at least one spine switch and multiple down-ports connecting to endpoints of the network; and
forwarding circuitry, which is configured to receive a multicast packet that is to be forwarded via a plurality of the ports, including the up-port and one or more of the down-ports, and to forward the multicast packet to the one or more of the down-ports using a normal scheduling process having a first forwarding latency while forwarding the multicast packet to the up-port using an accelerated scheduling process having a second forwarding latency that is smaller than the first forwarding latency,
wherein the forwarding circuitry is configured to store an entry in a cache memory identifying the up-port as having a highest path latency among paths through the network over which the multicast packet is to be forwarded, and to select the up-port for application of the accelerated scheduling process responsively to looking up the entry in the cache memory.

2. The network according to claim 1 , wherein the multi-layer topology comprises a Fat Tree.

3. The network according to claim 1 , wherein the multi-layer topology comprises at least one intermediate layer between the leaf layer and the spine layer, and the up-port connects the first leaf switch to the at least one spine switch through at least one of the switches in the at least one intermediate layer.

4. The network according to claim 1 , wherein the at least one spine switch is configured, upon receiving the multicast packet from the up-port of the first leaf switch, to forward the multicast packet to the second leaf switch using the accelerated scheduling process.

5. The network according to claim 1 , wherein the entry specifies at least a multicast flow to which the multicast packet belongs, and an identity of the path identified as having the highest path latency.

6. A packet switch for deployment as a first leaf switch in a packet communication network, which includes multiple packet switches connected together in a multi-layer topology, including at least a leaf layer, containing at least the first leaf switch and a second leaf switch, and a spine layer, containing at least one spine switch, through which the first and second leaf switches are interconnected, the packet switch comprising:
multiple ports, which are configured to serve as network interfaces for exchanging packets with the network and include at least an up-port leading to the at least one spine switch and multiple down-ports connecting to endpoints of the network; and
forwarding circuitry, which is configured to receive a multicast packet that is to be forwarded via a plurality of the ports, including the up-port and one or more of the down-ports, and to forward the multicast packet to the one or more of the down-ports using a normal scheduling process having a first forwarding latency while forwarding the multicast packet to the up-port using an accelerated scheduling process having a second forwarding latency that is smaller than the first forwarding latency,
wherein the forwarding circuitry is configured to store an entry in a cache memory identifying the up-port as having a highest path latency among paths through the network over which the multicast packet is to be forwarded, and to select the up-port for application of the accelerated scheduling process responsively to looking up the entry in the cache memory.

7. The switch according to claim 6 , wherein the entry specifies at least a multicast flow to which the multicast packet belongs, and an identity of the path identified as having the highest path latency.

8. A method for communication, comprising:
connecting multiple packet switches together in a network having a multi-layer topology, which includes at least a leaf layer, containing at least first and second leaf switches, and a spine layer, containing at least one spine switch, through which the first and second leaf switches are interconnected, wherein the first leaf switch comprises at least an up-port leading to the at least one spine switch and multiple down-ports connecting to endpoints of the network; and
receiving in the first leaf switch a multicast packet that is to be forwarded via a plurality of the ports, including the up-port and one or more of the down-ports;
forwarding the multicast packet to the one or more of the down-ports using a normal scheduling process having a first forwarding latency; and
forwarding the multicast packet to the up-port using an accelerated scheduling process having a second forwarding latency that is smaller than the first forwarding latency,
wherein forwarding the multicast packet to the up-port comprises storing an entry in a cache memory identifying the up-port as having a highest path latency among paths through the network over which the multicast packet is to be forwarded, and selecting the up-port for application of the accelerated scheduling process responsively to looking up the entry in the cache memory.

9. The method according to claim 8 , wherein the multi-layer topology comprises a Fat Tree.

10. The method according to claim 8 , wherein the multi-layer topology comprises at least one intermediate layer between the leaf layer and the spine layer, and the up-port connects the first leaf switch to the at least one spine switch through at least one of the switches in the at least one intermediate layer.

11. The method according to claim 8 , and comprising, upon receiving the multicast packet in the at least one spine switch from the up-port of the first leaf switch, forwarding the multicast packet from the at least one spine switch to the second leaf switch using the accelerated scheduling process.

12. The method according to claim 8 , wherein the entry specifies at least a multicast flow to which the multicast packet belongs, and an identity of the path identified as having the highest path latency.

Assignees

• Mellanox Tech Tlv Ltd

Inventors

• Levi Lion
• Marelli Amiad
• Elias George
• Zemer Oded
• Benros Yoav

Classifications

• H04L49/201

  Multicast operation; Broadcast operation · CPC title

• H04L49/25

  Routing or path finding in a switch fabric · CPC title

• H04L47/56 Primary

  implementing delay-aware scheduling · CPC title

• H04L12/1881 Primary

  with schedule organisation, e.g. priority, sequence management · CPC title

• H04L12/1877

  Measures taken prior to transmission · CPC title