06-07-2017 03:20 PM - edited 03-08-2019 10:53 AM
Hello;
According to Cisco documentation found on http://www.cisco.com/c/en/us/products/collateral/switches/nexus-7000-10-slot-switch/Data_Sheet_C78-437757.html. The Nexus 7000 with M series cards uses linear processing pooling the resources of all of the M series linecards together. I'm trying to find to acquire more information about that.
Let's say for example we have 4 x N7K-M224XP-23L on a N7K-C7009 chassis. Each of the cards do 120Mpps Layer3. If the scenario occurs were Linecard#1 is receiving traffics above 120Mpps, does this mean the card will forward traffic to other linecards on chassis for processing? and if it IS indeed doing that, how would the system handle VDC? Does it limit the forwarding power based on the linecards associated with the VDC? How about if two VDCs created splitting the ports of all 4 linecards, does this mean the each will have a hard limit of 240Mpps?
06-12-2017 06:53 PM
Hi,
I wasn't able to find any good document that can explain it but I don't think if a line card receives more than 120Mpps,it will forward it back to the fabric module and fabric module send it to other line cards that is not as busy. I think the limitation is 120Mpps, and if the card receives more packets than that, it will queue or drops them. And I think the same thing applies to cards that have ports in different VDCs.
HTH
06-12-2017 08:00 PM
Reza;
According to Cisco documentation about M1/2 linecards:
"The Cisco Nexus 7000 M2-Series module contains two integrated forwarding engines capable of delivering up to 120 million packets per second (Mpps) of Layer 2 and Layer 3 IPv4 unicast forwarding or 60 Mpps of IPv6 unicast forwarding across all ports of a single I/O module. The distributed architecture, with the forwarding engine integrated into each module, scales the forwarding performance of the chassis linearly by the number of I/O modules employed. The 18-slot chassis with 16 Cisco Nexus 7000 M2-Series modules can deliver up to 1.92 billion packets per second (Bpps) of IPv4 unicast forwarding or 960 Mpps of IPv6 unicast forwarding. Multicast forwarding is built into the I/O module, performing egress replication. The integrated forwarding engines also deliver ACL filtering, marking, rate limiting, and NetFlow with no degradation of performance. Powerful ACL processing supports up to 128K entries per module, where entries can address Layer 2, 3, and 4 fields in addition to the Cisco metadata fields that employ Security Group Tags (SGTs)."
":The Cisco Nexus 7000 M2-Series module contains two integrated forwarding engines capable of delivering up to 120 million packets per second (Mpps) of Layer 2 and Layer 3 IPv4 unicast forwarding or 60 Mpps of IPv6 unicast forwarding across all ports of a single I/O module. The distributed architecture, with the forwarding engine integrated into each module, scales the forwarding performance of the chassis linearly by the number of I/O modules employed. The 18-slot chassis with 16 Cisco Nexus 7000 M2-Series modules can deliver up to 1.92 billion packets per second (Bpps) of IPv4 unicast forwarding or 960 Mpps of IPv6 unicast forwarding. Multicast forwarding is built into the I/O module, performing egress replication. The integrated forwarding engines also deliver ACL filtering, marking, rate limiting, and NetFlow with no degradation of performance. Powerful ACL processing supports up to 128K entries per module, where entries can address Layer 2, 3, and 4 fields in addition to the Cisco metadata fields that employ Security Group Tags (SGTs).
The Cisco Nexus 7000 M2-Series module buffers data in Virtual Output Queues (VOQ) before the data flows to the fabric. The data flow is controlled by a central arbiter on the supervisor module, using a credit-based buffer design. This architecture offers a lossless fabric that delivers quality of service (QoS) and fairness across all ports, even during congestion."
Reference: http://www.cisco.com/c/en/us/products/collateral/switches/nexus-7000-series-switches/data_sheet_c78-706775.html
For example with the F2e linecards which deploy the SoC (i.e N7K-F248XP-25E) it does not make any reference to distributing linearly on chassis with the amount of I/O models installed. This is obviously because of the SoC design where each 4 ports share a single SoC.
Reference: http://www.cisco.com/c/en/us/products/collateral/switches/nexus-7000-series-switches/data_sheet_c78-720322.html
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide