Building Highly Resilient Networks in Network Edge - Part 2 - Device Link Groups

[MasonHarris]

This is the second part of a series of posts highlighting the best practices for customers who require highly resilient networks in Network Edge. This entry highlights how to build resilient Device Link Group (DLG) connections. For more information on Device Link Resiliency please refer to the Equinix docs page here.

• Device Link Groups are used to connect Network Edge device interfaces to other Network Edge device interfaces​
• They can be local (same metro) or remote and are a shared broadcast domain​
• A single DLG, like a single Ethernet cable, provides no resiliency​
• Local DLGs do not traverse the Equinix Fabric and aren’t susceptible to Fabric switch outages ​
• Hybrid DLGs have some devices connected to the Primary Fabric plane and other to the Secondary​
• Customers that require maximum resiliency should deploy additional DLGs that connect to both the Primary and Secondary Fabric networks​
• Redundant Devices can have one or both devices connect to the DLG​
• Clustered devices act as one logical device, therefore the cluster is represented as a single entity​

Each device will build connections to the Plane to which it is connected as shown here:

• Redundant Devices show as individual entities in the DLG workflow​
• Clustered Devices are represented as a single logical entity, connecting the DLG to the active node (node 0 is active by default)​

• In October 2023 a new feature was released that allows customers to choose which Fabric plane the DLG is connected to: Primary or Secondary​
• Previously, all connections were made over the Primary Fabric plane by default​
• Customers now have control over the level of resiliency for their DLGs. Single DLGs will have no resiliency but customers may opt to deploy a 2nd, parallel DLG for connections that are business-critical.​
• The 2nd DLG will incur additional Fabric charges and will consume an additional interface on every connected Network Edge device​
• We do not see value in creating multiple, local (same metro) DLGs for resiliency, but we do not restrict customers who want to deploy in this manner​
• Local DLGs do not traverse the Fabric switches and incur no charges today​
• Customers now have the control to deploy resiliency with DLGs based on their requirements and budget as shown below​

The next post will cover the best practices for creating resilient EVP-LAN connections. You can read the first part of this series here.

[MasonHarris]

The next post will cover the best practices for creating resilient EVP-LAN connections. You can read the first part of this series here.

[marek_moszynski]

Thank you @MasonHarris for this detailed presentation!

When you say "Local DLGs do not traverse the Equinix Fabric and aren’t susceptible to Fabric switch outages ", is there an outage scenario where VNF1 and VNF2 connected by a Local DLG would stay up, but the Local DLG would go down? In other words, does a Local DLG traverse a piece of infrastructure independent from VNF1 or VNF2?

[MasonHarris]

Hi @marek_moszynski thanks for the question--i can't think of a scenario where VNF1 and VNF2 in the same metro connected via a local DLG would be up and the DLG would be down. At least in my knowledge this has never occurred. The point I was trying to make it that local DLGs (same metro) don't go out to the Fabric (Primary or Secondary) and therefore are not susceptible to the same resiliency scenarios as remote DLGs, EVP-LANs and VCs. We do get customers asking if they should build redundant local DLGs and, while we don't restrict it (and there is no charge), we don't see the benefit as the cabling between the computes is already redundant and in the same cage. I hope that clarifies please let me know if you have more questions.

[marek_moszynski]

crystal clear, thanks for the confirmation!