Cisco ACI Fabric Topology High Level Overview

• ACI uses Spine-Leaf topology (Architecture) or Clos topology — https://medium.com/@ajaykul/data-center-networking-part-i-511b6cfcc8fe
• Spine ⇒ Cisco Nexus N9K Spine switches.
• Leaf ⇒ Cisco Nexus N9K Leaf switches
• APIC controllers apic-application-policy-infrastructure-controller.md uses minimum 3 and they are in a cluster.
  • Each APIC is connected to two leafs for redundancy.
  • 3 x APIC because of Split brain or Quorum. In other words, the 3rd APIC acts as a tiebreaker.
  • APIC controller are Cisco UCS x86 servers.
• ACI is especially great for east<->west traffic compared to North<->South traffic in the traditional DC.

Scalability

* If you need more _**ports**_ for the endpoints (VMs, BareMetal servers, Storage) simply add more leaf switches.

* If you need more _**bandwidth**,_ simply add more spine switches.

Traditional networking vs ACI Policy Based Networking

#### Traditional networking * Easy to set up initially. * Harder to scale. * More prone to user error (ie: Sys admin configuring errors) due to manual work. * Harder to track application level (granular even) traffic. * It is harder to troubleshoot network errors. #### ACI Policy Based Networking * Harder to set up initially. * It gets easier to scale up. * Easier to catch network errors and troubleshoot. * Less prone to user error due to the APIC controllers being responsible to push config to the network devices. The user simply configures the policy. * Traffic analysis to the application level is a breeze.