Ask the Expert: Cisco Data Center Overlays with Focus on VXLAN

[ciscomoderator]

Welcome to this Cisco Support Community Ask the Expert conversation.  Cisco Data Center Overlays with focus on VXLAN (Virtual Extensible Local Area Network) to  Vishal Metha.

Ask questions from Tuesday October 20, to Friday October 30, 2015

In the modern data center, traditional technologies are limiting the speed, flexibility, scalability, and manageability of application deployments. There is emerging interest in the industry in overlay technologies (such as MPLS, VXLAN, LISP...) which may address some of these challenges. This session will discuss the latest trend in Overlay Technologies and compare each one of the available solutions in terms of deployment, benefits and challenges. The major emphasis in this session will be on VXLAN evolution and to elaborate on which Cisco Products support different VXLAN solutions. Various VXLAN configurations such as BGP-EVPN control plane, L2 Gateway, L3 Gateway; Bridging, Packet-Flow, Best Practices, Deployment scenarios and other advanced features will be also discussed.

Vishal and Pranav  will be helping you with all your queries on all of the above.

Vishal Mehta is a Technical Marketing Engineer with Cisco's Data Center Competitive Insights Team based in San Jose, California. Previously he was working as the customer support engineer for Data Center Server Virtualization Technical Assistance Center (TAC) team for the past 4 years with a primary focus on data center technologies such as Cisco Nexus® 5000, Cisco UCS, Cisco Nexus 1000V, and Virtualization. He has CCIE® certification (# 37139) in Routing/Switching, Service Provider & Data Center. Vishal has presented at Cisco Live in Orlando 2013, Milan 2014, and San Francisco 2014 (BRKCOM-3003, BRKDCT-3444, and LABDCT-2333).

Pranav Doshi is a Technical Marketing Engineer with Cisco’s Data Center Competitive Insights Team based in San Jose, California. Pranav has been with Cisco since 2010, starting off as a Team-Lead Technical Support Engineer in the LAN-Switching team working on Catalyst 37XX/45XX/65XX and Nexus 2K/3K/5K/6K/7K/9K platforms. After working in Cisco TAC, Pranav worked in Cisco Advanced Services as part of the Data Center Solutions Team for Cisco’s Enterprise Customers. During his stint with Cisco AS, Pranav designed data centers with Nexus Standalone deployments leveraging technologies like Fabricpath, OTV  as well as ACI deployments.

Find other  https://supportforums.cisco.com/expert-corner/events.

Because of the volume expected during this event, Vishal or Pranav  might not be able to answer every question. 

**Ratings Encourage Participation! **
Please be sure to rate the Answers to Questions

[Vishal Mehta]

Hello,

Sure you have all valid questions and my answers are based from design guide:

http://www.cisco.com/c/en/us/products/collateral/switches/nexus-9000-series-switches/guide-c07-734107.html

Please find your requested answers below:

1. In vPC VTEP scenario with MP-BGP EVPN VXLAN do we need to have the vPC peers to be BGP adjacent? if YES, why we need this adjacency?

Answer: No, we dont need vPC peers to form BGP adjaceny amongst each other

2. Can you explain why we need VRF overlay VLAN and SVI for VXLAN routing?

Answer: VRF allows to create multiple tenants which can have over-lapping address

3. Why VRF overlay SVI does NOT have an IP address assigned?

Answer: Because for VXLAN routing, the local VTEP needs to rewrite the inner destination MAC address to that of remote VTEP's router MAC Address.

4. DCI between two DCs with a pair of vPC VTEPs in each. The DCs are interconnected through a multihop MP-eBGP.

  4.1 can you propose me a documentation regarding the this design?

  4.2 is the full-mesh eBGP between the four VTEPs is the proper solution?

  4.3 should I distribute the MP-BGP EVPN routes into inter-DC IGP?

  4.4 how I must tune the MP-BGP to use both opposite VTEPs for load balancing?

Answer: VXLAN Control Plane for DCI currently can be considered as work in-progress. It was not originally meant for this solution but it has all bits-n-pieces to be used as DCI.

There is good blog post on this topic by Yves - http://yves-louis.com/DCI/?p=965 It covers scenarios and packet walks as well.

For details on specific DCI design and best-practises, i would recommend to reach out to your local Cisco Representative or Network Consulting Engineer who can help you to design DCI-VXLAN as per your network requirements.

Thanks,

Vishal

[stanibarb]

thank you very much Vishal. It helps me a lot!

[anthonypriore]

Hello

I am currently looking for information on Vxlan as a DCI specifically DCI between two DCs with a pair of vPC VTEPs in each. Would you be able to provide me with a good design guide and possible configuration example and this type of setup.

Thanks So Much

Anthony

[raugusti]

Hi gents,

I hope my question is not out of the scope, as it is related to Nexus switches and VXLANs: customer decided to use NSX and part of the design includes Cisco spine-and-leaf switches. However, we can't agree if we should use NX-OS (with N93936 as leafs and N9508 as spine switches) or ACI approach. Can we even provide vPC functionality if first scenario is used (one with NX-OS)?

Thanks,

Renato

[Vishal Mehta]

Hello Renato,

Its very fair question to ask. Yes you can provide vPC functionality with Nexus 9000 switches using VXLAN. Below link gives more details on vPC deployment on VXLAN

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/6-x/vxlan/configuration/guide/b_Cisco_Nexus_9000_Series_NX-OS_VXLAN_Configuration_Guide/b_Cisco_Nexus_9000_Series_NX-OS_VXLAN_Configuration_Guide_chapter_010.html#concept_C769B7878...

Thanks,

Vishal

[raugusti]

Thanks for your swift response Vishal!

Here is the detail that bothers me: vPC requires two neighboring switches to be connected with peer-link but spine-and-leaf topology doesn't allow two leafs to be directly connected, right? I'm talking here about N9ks running NX-OS (not ACI) and I even found one article (link below) where spine-and-leaf topology is described but  vPC is not mentioned at all. So, I'm looking for your help to clarify if vPC will work in that scenario (N9ks running NX-OS in spine-and-leaf topology)?

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/6-x/fundamentals/configuration/guide/b_Cisco_Nexus_9000_Series_NX-OS_Fundamentals_Configuration_Guide/b_Cisco_Nexus_9000_Series_NX-OS_Fundamentals_Configuration_Guide_chapter_01.ht...

[Vishal Mehta]

Hello Renato,

That is very good question - Yes Leaf-Spine Topology design doesn't allow two leafs to be connected with each other. But vPC on Leaf-Spine fabric is special design and it is supported on Nexus 9ks (standalone mode).

The guide you are referring is basic configuration on Nexus 9k platform for features like POAP, Terminal settings, File Systems and so on..

Please refer to link below on VXLAN-vPC Configuration:

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/7-x/vxlan/configuration/guide/b_Cisco_Nexus_9000_Series_NX-OS_VXLAN_Configuration_Guide_7x/b_Cisco_Nexus_9000_Series_NX-OS_VXLAN_Configuration_Guide_7x_appendix_0101.html

Thanks,

Vishal