Leaf & Spine: L3 Vs. L2 (VLT)

A short & sharp comparison of Layer 3/ECMP Leaf & Spine vs. Layer 2 Multi-chassis Link Aggregation, e.g. Dell EMC VLT (Virtual Link Trunking) Topologies.

Leaf Spine vs VLT

Spine & Leaf (Layer 3) VLT/vPC/MCLAG (Layer 2)
Interoperability Built on Open/standards based Routing Protocols.

In theory, each layer can be multivendor as long as it supports ECMP/Routing Protocol of choice.

Proprietary. Lock-in to the respective vendor.
Scalability in the Core Very scalable. Total limited only by number of spine ports, oversubscription desired, & number of interconnections between them. Limited to 2.
Traffic Pattern East West optimized With SVIs in Core, Flatter/bigger broadcast domains. Could be identical to SL
Core Congestion Less impact due to flatter design More impact as tiered/hierarchy.

ARP response and L2/L3 protocol Scale handled by Core.

Latency All devices at equal hops from each other, so predictable latency Depends. Can be identical to SL.
Routing/Switching ECMP. Predictable. Load balancing.

Additional IP Address complexity for P2P links as these are L3

Protocol choice could increase Core overhead with scale (OSPF/Link State)

Layer 2 Port Channel LAG groups.  Active Active uplinks

No ip address complexity for P2P links as these are L2.

FHRP/Gateway Redundancy In Layer 3 Spine/Leaf, FHRP in Access Layer/ Leaf ToR FHRP in Core/Agg Typically.

(still possible to use a model with routed access layer)

When using of FHRP/GW redundancy, ensure L2/L3 roots are aligned.

L2 Fault Domain (VLAN/STP) In Layer 3 Spine/Leaf, Local to Rack. Stops at Rack ToR pair. When VLAN/Broadcast domain SVIs in the Core, it Spans the Network.
Host Mobility + Multitenancy No, unless Overlay/Encapsulation like VXLAN Yes
Upgrade path Pay as you grow

  • More Access ports = Add ToR Leaf Switches
  • More bandwidth = Add Core Spine Switches
Forklift – higher capacity/capability switches
Cabling Complexity Significant, and increases with the number of units Consistent, minimal.
Maintenance No downtime No downtime
Failure Scenarios Resilient layers. Can sustain multiple unit outages. Split Brain possible for geographically dispersed Pairs. If manifested, would severely impact routing/forwarding.
Ideal Use Scale out for both bandwidth and edge ports. (spine & Leaf layers), Layer 3 Multi-path Simplicity, Layer 2 Multi-pathing with Active Active LAG group & Dual Active Control planes.

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s