Spine and Leaf Architecture

Lessons Discussion
1

This topic is to discuss the following lesson:

2

Hi Rene, excellent post. I have a question regarding vPC on your Conclusions, “vPC somehow solves this limitation of STP, but vPC can only use two active links.” - what do you mean by “only use two active links”?

3

Hello Laura

When configuring vPC peers, you can only connect to two devices. Specifically, Cisco states:

You can have only two devices as vPC peers; each device can serve as a vPC peer to only one other vPC peer. The vPC peer devices can also have non-vPC links to other devices.

Note some invalid configurations below:
image
You can find out more info at this Cisco documentation:

So vPC provides redundancy in pairs, and does not even come close to the level of redundancy made available by the spine and leaf architecture.

I hope this has been helpful!

Laz

4

Hi guys
I think you don’t get Laura’s point- you can only connect 1 switch to 2 vpc peers , but that does not mean you only can have 1 uplink to each peer.
We have in our datacenter configured some switches with 8 uplinks to each 5k nexus, so you have 8 links participating in 1 vpc
The picture you showed is for 1 switch connecting to 3 vpc peers - that is not possible, but 1 device connected to 2 vpc peers using 8 links on switch and 4 on each vpc peer is doable and frequently used

5

Hello Marek

Yes, thanks for the clarification. It is true that you can have each individual uplink to a peer be composed of an etherchannel, so that you can have up to 8 physical links to each of the vPC peers. That way you can have up to 16 physical links.

However, the limitation that Rene was speaking about in the lesson is the fact that you are limited to uplinks to a maximum of two vPC peers. You cannot have uplinks to more than two physical switches like you can with spine and leaf.

I hope this has been helpful!

Laz

6

Hi,
If we have spine and leaf architecture , where do we place firewall ?
Thanks

7

Hello Sims

For the spine and leaf model, you must remember that the spine switches are only connected to leaf switches, and nothing else. The spine is the backbone. The leaves are connected to all other devices including servers, firewalls, load balancers, and edge routers. This allows all devices to be exactly the same number of segments away from each other.

So not only are firewalls connected to the leaf switches, but everything is connected to leaf switches.

I hope this has been helpful!

Laz

8

are there any lessons about VXLAN?

9

Hello Giovanni

There is some information in the following lesson:

However, a more detailed lesson will be included in the CCIE Enterprise course which @rene is currently working on. It should be a matter of weeks, we’ll keep you posted.

I hope this has been helpful!

Laz

1 Like
10

Hi,

Can you provide an simple example about a vpc configuration between 1 device and 2 switches?

Thanks

11

Hello Giovanni

vPCs allow links that are physically connected to two different switches to appear to a third downstream device to be coming from a single device and as part of a single PortChannel. The third device can be a switch, a server, or any other networking device that supports IEEE 802.3ad PortChannels.

That third device, be it a server or another switch, does not actually participate in the vPC procedure. It is configured using etherchannel as it would be as if its physical links were physically connected to the same switch. The vPC peer link involves only the link between the two switches.

You can see a basic vPC configuration between two switches on page 65 of the following Cisco documentation.

The configuration on the side of the server will be the same as that used to configure any normal etherchannel configuration. You can also find out more about vPC configuration best practices here.

I hope this has been helpful!

Laz

12

Does Rene plan a lesson about that?
At my opinion this material is a bit complex to understand, based only on cisco documentation.

13

Hello Giovanni

vPCs are exclusively a feature of the Cisco Nexus series of devices. Cisco IOS doesn’t support it, and since most certification material involves Cisco IOS, it is not included as part of the specific Cisco certifications. Certifications such as the new CCNP Datacenter certification may include Nexus-based features, but this certification is not yet covered in Networklessons.

If you’re interested in having this covered in the future, feel free to make a suggestion at the Member Ideas page below:

Cisco IOS does however support etherchannel connections to two physically different switches using other features. One way is to configure a stack of switches using Stackwise, and create etherchannels using ports on multiple switches. Another way is to use the VSS feature on higher end chassis switches where two devices function logically as one.

Both of these technologies make multiple physical devices function as a single logical device, thus allowing an etherchannel link to span multiple physical switches. But just to clarify, this is a different technology than vPC which is supported only by Nexus series devices.

I hope this has been helpful!

Laz

14

Hi,

First of all thank you for your explanation, but i’m wondering is the Spine and Leaf Architecture in the encore exam? I dont see this anywhere on the Cisco 350-401 encore exam description

15

Hello Sven

Within the Cisco ENCOR blueprint, the “spine and leaf” terms are not found. However, because Cisco tends to write the blueprint in a relatively vague way, it may not be mentioned by name. As is also mentioned in the official Cisco ENCOR exam topics page:

The following topics are general guidelines for the content likely to be included on the exam. However, other related topics may also appear on any specific delivery of the exam.

The content in NetworkLessons is decided upon based on the blueprint, the official Cisco certification guides as well as experience from the real certification exams themselves. We do our best to include everything that will be useful to you in your preparation. For this reason, it has been determined that this may be a topic that does appear in one of the exams. Specifically, it fits well within the following exam topics:

  • 1.1 Explain the different design principles used in an enterprise network
  • 3.0 Infrastructure

I hope this has been helpful!

Laz

2 Likes
16

Hi Rene/Laz,

You mentioned that one of the disadvantages of the three-layer hierarchical is: ‘‘Limited bandwidth, vPCs solve the STP problem that we can only use one active link, but vPCs are limited to two active uplinks’’, what about configuring ‘‘etherchannels’’ on all the up-link ports?

17

Hello Walter

When you configure etherchannel, you are increasing the bandwidth made available and you are bypassing STP, but only for links between the same two switches. If you create etherchannel links to multiple switches, all but one of those etherchannel bundles will be blocked by STP in order to avoid L2 loops. Whether we are blocking physical links, or etherchannel bundles, the result is the same.

Even if you create etherchannels that span multiple switches, you can only span them across two switches, using vPCs, so the same limitation applies.

I hope this has been helpful!

Laz

18

Hi Rene,

Can you help me what is the meaning of Fabric in Data center ? what is POD ?
Can you give me a example how to use L3 between spine and leaf ?
what is 3 tier , 2 tier and 1 tier architecture in Spine and Leaf ?

19

Hello Ravi

In order to cover the topics that you are asking for, you will have to do a bit of reading. Hopefully, however, I can help you by pointing you in the right direction.

In a data center, the fabric is a term used to describe the network topology of that datacenter. However, because data centers use networking technologies in addition to traditional physical connectivity, the fabric includes these technologies as well. So the fabric typically includes computing, networking, and software components that work together in the data center to deliver the network services.

Now this fabric must be designed in a modular way, meaning it must be scalable, and it must have components that can be added or removed as whole entities. One of these entities is called a pod. A pod is a portion of the datacenter that comprises interconnectivity for a particular set of servers and/or services. You can see an example of a pod in the diagrams in the lesson. A Pod can be a component of the traditional 3-tier architecture, or it can also be a part of the more modern data center fabric architecture.

Using L3 simply means that all connections between your spine and leaves are routed connections using IP addresses in different subnets.

The tier architectures are different than spine and leaf. Spine and leaf does not have a tiered structure. If you go through the lesson, you will see how the differences between these architectures are described in detail.

I hope this has been helpful!

Laz

20

i have gone through the other docs, but got confused on the terms.
Now it clears to me.Thanks Lazaros

1 Like