OTV: Overlay Transport Virtualization

OTV(Overlay Transport Virtualization) is a technology that provide layer2 extension capabilities between different data centers.
In its most simplest form OTV is a new DCI (Data Center Interconnect) technology that routes MAC-based information by encapsulating traffic in normal IP packets for transit.

• Transparent workload mobility
• Business resiliency
• Superior computing resource efficiencies

References:

Cisco: OTV Quick Start Guide

Cisco: NX-OS OTV Configuration Guide

Cisco: OTV Best Practices

Cisco: OTV Whitepaper

OTV Encapsulation

OTV adds a further 42 bytes on all packets traveling across the overlay network. The OTV Edge device removes the CRC and 802.1Q fields from the original Layer2 frame. It then adds an OTV Shim Header which includes this 802.1Q field (this includes the priority P-bit value) and the Overlay ID information. It also includes an external IP header for the transport network. All OTV packets have Don’t Fragment (DF) bit set to 1 in the external IP header.

(more…)

https://www.cisco.com/c/dam/en/us/solutions/collateral/data-center-virtualization/data-center-interconnect/at_a_glance_c45-493703.pdf

• Types of Data Center Interconnections
  • Layer 1 DCI
    • Physical connection between Data Centers
      • Dark Fiber ( MetroE)
      • Fibrechannel
    • Not as flexible
  • Layer 2 DCI
    • Carrier appears as L2 Bridge
    • MPLS options
    • VPLS options
    • IP Mobility
  • Layer 3 DCI
    • IP Based connection
      • iSCSI
      • Routing
      • ATOMoGRE

(more…)

Software Defined Networking

• Advantages SDN
  • Automatic Infrastructure Provisioning
  • Multi-tenant enviornments
  • Flexible Placement of servers ( Mobility )
  • Health monitoring of applications
  • Application to NET ( Southbound ) and NET to application ( Northbound ) communication
• Cisco’s SDN implementation: Application Centric Infrastructure ( ACI )

Three key ingredients for ACI

• Nexus 9000 series / 9300 / 9500.
• Aplication Policy Infrastructure Controller ( APIC ).
  • Cisco recommends a minimum of three APIC servers.
• Policy Model ( “What talks to what and how” ).

(more…)

TRansparent Interconnection of Lots of Links

• https://en.wikipedia.org/wiki/TRILL_(computing)
• Often in datacenters a layer 2 design is implemented to span VLANs.
• First define your FabricPath domain, which disables Ethernet.
• IS-IS enabled Fabric Path Domain.
• Classic Switch data learned and IS-IS Level 2 ‘Routing Table’ Built.
• Opens L2 ECMP Capabilities.
• Leaf to servers with vPC / MEC

(more…)

What is a vPC (virtual Port Channel)

• Nexus series Network Virtualisation Technology.
• “Lightweight” VSS – Combine ports, not switches.
• Links on different switches to appear as the same device.
• Downstream device can be anything supoprting 802.3ad (LACP).
• Commonly called Multi Chassis Etherchannel ( MEC ).

(more…)

STP in the datacenter

STP logical interfaces maximum

• RSTP / MST reach topology maximus. Switch specs logical / virtual interfaces.  (6500/ 6748)
  • Logical interfaces == ((trunk ports * vlans ) + non-trunk interfaces)
  • Virtual interfaces ( Per line card ) = (( trunk ports * vlans ))
  • verify with ‘show spantree sum total‘

Example logical interfaces:

6500 Chassis:

• 120 VLANS
• 49 Access layer switches
• 2 connections to each switch ( double uplink but with etherchannel counts as 1 logical interface )
• 1 Cross Connect to secondary Agg Switch.
• 30 other devices connected
  • (( 120 * 50 ) + 30 ) = 6030 Logical interface (out of 10,000 max)

Example Virtual interfaces:

Cisco 6748 Linecard

• 120 VLANS
• 12 Access layer switches, 4 Etherchannel-bundled connections.
  • 12 x 4 = 48 ports  ( Virtual interfaces counts every interface )
    • ( 120 * 48 ) = 5760 Virtual Interfaces (out of 1800 max)

High Availability

Nic teaming options

• Adapter fault tolerance (AFT):  active / standby  – Two nics  One switch
• Switch fault tolerance (SFT):  active / standby  – Two nics Two switches
• Adaptive Load Balancing (ALB):  active / active – One IP Two Macs
• Etherchannel – LAG

Expectations and Failover Timers

Blade Server design

• Overleg with the server team connectivity.
• Many blade servers enter de “enterprise switch” market with an integrated switch.
• Pass-through cabling or integrated switches.
• Significant impact on power – cooling – weight.

Connecting the blade to the network

• If you use the integrated blade switch, use a layer3 access layer.
• Avoid a double Layer 2 design:
  •  [Layer 2 on the access layer] connected to a [layer 2 domain within the bladeswitch].
• If you use passthrough, use a layer2 or layer3 access layer.

Scaling the Datacenter Architecture

(more…)

Core

• Not all datacenter designs needs a core layer
• Access to aggregation, aggregation to core : 10 or 40GBps
• CEF load balancing tuning (L3 + L4)
• Core should run L3 only, Aggregation acts as L3/L2 boundry to access
• Core runs OSPF / EIGRP with aggregation

Aggregration

(more…)