Cisco Catalyst 9300: StackWise vs VSS Explained - 夜莺博客

Cisco Catalyst 9300: StackWise vs VSS Explained

Ask two engineers whether a Catalyst 9300 can run VSS and you will get two confident, opposite answers. The useful answer is that the 9300 uses StackWise — in three throughput tiers, plus StackWise Virtual for two-chassis pairs — while classic VSS is a legacy virtual-switching technology aimed at chassis and older fixed platforms where two devices are joined over ordinary front-panel links. This article untangles the family, compares the throughput numbers that actually matter, and gives a decision rule for access versus core/distribution.

Three different mechanisms, one goal

All three present multiple physical devices as one logical switch to the network, so the downstream devices see a single LACP partner, a single control plane and a single spanning-tree root:

Mechanism Interconnect Typical scale Platform focus
StackWise (9300 backplane) Dedicated stacking ports and cables Up to 8 (16 on some models/code) Access layer
StackWise Virtual Front-panel 40G/100G links plus a keepalive Exactly 2 chassis Distribution / collapsed core
VSS Front-panel links (EtherChannel) plus VSL 2 chassis Legacy 6500/4500-X class platforms

StackWise throughput tiers on the 9300 family

StackWise throughput is not one number. Cisco's architecture white paper documents three tiers with different ring counts and per-stack bandwidth:

Version Rings Throughput per ring Full-ring stack throughput With SRP
StackWise-1T (9300X modular uplink) 6 90 Gbps 540 Gbps 1 Tbps
StackWise-480 (9300 modular uplink) 6 40 Gbps 240 Gbps 480 Gbps
StackWise-320 (9300 fixed uplink) 4 40 Gbps 160 Gbps 320 Gbps

The "with SRP" column is the number that appears in marketing material; the plain full-ring figure is the sustained figure to design against. Mixed stacking between 9300X and 9300 SKUs is supported, which lets a single logical switch span a capacity upgrade without a forklift — the stack simply runs at the lowest common tier.

Why StackWise is the access-layer answer

Backplane stacking wins on port density and distance-free cable management: the stack cables are short, non-network-grade links between physically adjacent switches in one closet, so there is no optics cost and no possibility of a user unplugging the stack link. 8-16 member stacks give very large logical access switches with a single management address, and because the stack shares one control plane, NIC teaming on servers can use active-active LACP rather than active-standby.

The trade-off is physical locality: members must be close enough for the stack cables, and a stack is a single failure domain in which an IOS upgrade reboots the whole logical switch. Cisco's StackWise to support mixed generations, but stack-wide reloads remain the operational cost.

When StackWise Virtual replaces VSS

StackWise Virtual joins two 9300s over front-panel 40G/100G ports rather than a backplane. That makes it the modern equivalent of VSS for a distribution or collapsed-core pair: the two chassis can sit in different racks, up to the optical reach of the links you provision. The architectural differences that matter:

  • Scale: two members only, versus 8-16 for backplane stacking.
  • Interconnect: front-panel ports consumed for the virtual link, plus a separate Layer 3 keepalive path — so budget ports and optics rather than stacking cables.
  • Dual-active detection: uses the same family of mechanisms as VSS (PAGP, BFD or an enhanced PAgP over a redundant path) and must be configured deliberately; without it a split keeps both chassis active and duplicates the control plane.
  • Topology position: because the pair is physically separated, StackWise Virtual pairs are usually the distribution layer, with access switches dual-homed below them.

VSS itself: still relevant, different platforms

VSS joins two switches with a Virtual Switch Link built over a standard EtherChannel of front-panel ports, elects an active and a standby chassis, and runs a single control plane. It is functionally the same idea as StackWise Virtual, and Cisco's own community guidance treats VSS and StackWise Virtual as the same concept applied to different hardware generations. Where you will still meet VSS in production is on older fixed 10G/40G platforms and legacy chassis at core/distribution, not on a 9300.

Decision rule

  1. Cabling endpoints in one or a few adjacent closets, want maximum port density per management address, tolerate stack-wide reloads → backplane StackWise.
  2. Need two physically separate chassis to share one control plane at distribution, with room for front-panel virtual-link ports and a keepalive path → StackWise Virtual.
  3. Running legacy chassis or older fixed platforms that only support the feature → VSS, with the same dual-active detection discipline.

Whatever the mechanism, the three operational items that decide whether the pair behaves are: a dedicated keepalive path on its own link or VLAN, dual-active detection enabled and tested, and identical configuration on both members — a single-chassis configuration drift is what turns a redundant design into an outage.

For the throughput-tier and VSS comparison in table form see StackWise-480 vs StackWise-1T vs VSS, and for the failure-handling side see StackWise Virtual dual-active detection.

原文链接:https://www.cisco.com/c/en/us/products/collateral/switches/catalyst-9300-series-switches/white-paper-c11-741468.html