Руководство По Проектированию для Cisco Cisco Nexus 5010 Switch
© 2012 Cisco and/or its affiliates. All rights reserved. This document is Cisco Public Information.
Page 17 of 20
Pathcost Method Long
If you are not using spanning-tree pathcost method long, the cost of most links in a data center is almost identical
regardless of the available bandwidth along a given path:
●
1-Gbps link = 4
●
2- Gbps links = 3
●
3- Gbps links = 3
●
4- Gbps links = 3
Considering that spanning tree was designed before the availability of 10 Gigabit Ethernet or even Gigabit
Ethernet, the default spanning-tree reference cost for a link is inadequate, as most links with 10 Gigabit Ethernet
bandwidth or bandwidth with multiples of 10 Gigabit Ethernet will appear to have the same cost.
With spanning-tree pathcost method long enabled, spanning tree calculates the best forwarding path according
to the link bandwidth to the root and not based on hop count. With this command, the cost of links is as follows
(these numbers are weights so they don’t have a specific measurement unit):
(these numbers are weights so they don’t have a specific measurement unit):
●
One Gigabit Ethernet link = 20,000
●
Two Gigabit Ethernet links = 10,000
●
Three Gigabit Ethernet links = 6660
●
Four Gigabit Ethernet links = 5000
The cost with 10 Gigabit Ethernet links is as follows:
●
One 10 Gigabit Ethernet link = 2,000
●
Two 10 Gigabit Ethernet links = 1,000
Scaling Considerations
Design scalability is based on various factors, including:
●
Number of PortChannels supported (discussed in Chapter 3 of this guide)
●
Number of VLANs supported by a switch
●
Logical interface count (also called bridge state count) for the control plane of a given switch
●
Oversubscription rate
Logical Interfaces or BPDU States
A logical port is the sum of the number of physical ports times the number of VLANs on each port. This value is
what produces load on the CPU because each port that includes a VLAN has to generate and process BPDUs.
The more VLANs on the trunks, the higher the load on the CPU. This load is one of the reasons you should clear
trunks from unnecessary VLANs.
The upper limit of logical ports that a switch can support depends on the spanning-tree algorithm. With Rapid
PVST+, each VLAN is a separate instance of the spanning-tree algorithm, which means that Rapid PVST+ can
support fewer VLANs than MST.