Networking & Content Delivery

Migrating AWS Direct Connect to a new location

As new AWS Direct Connect locations become available, we recommend customers review their options to make sure they are using the best route to connect to AWS. Many times, moving a connection to a Direct Connect location that is geographically closer to your data centers (DCs) and branch locations can improve network performance, and might reduce your last-mile connectivity costs. In this post, we cover key design considerations and walk through the steps for moving an existing Direct Connect connection to a new location.

Design considerations and best practices

Overview of the migration process

Migrating Direct Connect connections to a new location involves:

  1. Ordering a Direct Connect connection at a new location
  2. Configuring VIFs and on-premises devices
  3. Testing failover to confirm traffic flow over the new connection
  4. Shifting the traffic over to the new connection
  5. Decommissioning the old connection.

Our example in this post is moving a network connection from a location in Sydney, Australia to one in Auckland, New Zealand. However, the same steps apply to moving any Direct Connect connection, regardless of location and AWS Region.

Migration process step-by-step

At the start of our migration project, we have two Direct Connect connections at two separate Direct Connect locations, both in Sydney. In the following diagram (figure 1), we show the primary connection at an Equinix Direct Connect location using a solid blue line on the left side. The secondary connection to the NextDC Direct Connect location in Sydney is shown on the right with a dotted purple line.

Starting State: Two Direct Connect connections at two Direct Connect locations

Figure 1: Starting State: Two Direct Connect connections at two Direct Connect locations

The primary connection is preferred for outbound traffic (on-premises network to AWS) because the Local Preference on the routes (received from AWS) has been set to a higher value (300) as compared to secondary connection (Local Preference is set to 200).

The primary connection is also preferred for inbound traffic (AWS to the on-premises network) because we’re advertising on-premises routes to AWS with local preference BGP community tag set to 7224:7300 (high preference). While on the secondary connection, we’re advertising on-premises routes to AWS with local preference BGP community tag set to 7224:7200 (medium preference).

If you want more details on how to use BGP attributes to influence traffic over multiple Direct Connect connections, the knowledge center article How do I set an active/passive direct connect connection to AWS will be helpful.

Step 1: Setup a third Direct Connect connection at a new Direct Connect location

First, you must establish a new dedicated or hosted Direct Connect connection at the new Direct Connect location using the steps described on the Direct Connect Getting Started page.

Once the new Direct Connect connection is available in your AWS Management Console, use it to create a Private Virtual Interface (VIF) for testing the connectivity over new Direct Connect connection between on-premises and AWS. For the purpose of a simple connectivity test, you can terminate this VIF directly on a VGW attached to a test VPC where you have an EC2 instance with security groups allowing ICMP protocol from your on-premises network.

Complete the BGP configuration on your on-premises router to advertise and receive appropriate routes to test connectivity. Once the BGP status is showing “up” for this test VIF, use ping to verify that you can connect to the EC2 instance hosted inside your test VPC.

Now that you have confirmed that you can reach AWS from your on-premises network, you can delete this test VIF. Next, we’ll create a new VIF over this new Direct Connect connection for carrying your actual workload traffic.

Step 2: Setup virtual interfaces (VIF) on the new Direct Connect connection

We strongly recommend that you perform the subsequent steps during an agreed maintenance window.

Create a new VIF of the same type that you have on your primary and secondary Direct Connect connection. If your primary and secondary connections are terminated on a Direct Connect Gateway, terminate new VIF on the same Direct Connect Gateway.

Next use the following traffic engineering techniques to keep the new connection in standby. You can use BGP attributes to make sure production traffic is not sent to the new Direct Connect connection (represented in dashed black line in figure 2).

  • Traffic from your on-premises network to AWS 
    The route with the higher Local Preference value is preferred. Therefore, set the Local Preference for the routes received from AWS to a value that is lower (100) than the other two connections.
  • Traffic from AWS to your on-premises network
    To keep the new Direct Connect connection in standby mode, apply a local preference community tag with a lower preference (7224:7100) to all on-premises routes you are advertising to AWS. This ensure that any traffic originating from AWS continues to prefer existing Direct Connect connections.

Create a third Direct Connect connection at a new Direct Connect location

Figure 2: Create a third Direct Connect connection at a new Direct Connect location

At this stage, your Direct Connect setup should look as described in figure 2.

Step 3: Failover testing to test traffic over new VIFs

Follow the instructions on AWS Direct Connect Failover Test to gracefully shut down the BGP sessions over both primary and secondary Direct Connect connections. This will force any new traffic between your on-premises network and AWS over the new VIFs configured over the new Direct Connect connection (represented in dashed black line in figure 3).

Direct Connect failover test to force traffic over new connection

Figure 3: Direct Connect failover test to force traffic over new connection

Test your workloads on the new Direct Connect connection to make sure they’re working as intended and meeting your performance requirements. If you encounter a problem, you can stop the test immediately to bring up BGP peering sessions on the VIFs associated with primary and secondary Direct Connect connections and reinstate the flow of traffic in its original state.

Step 4: Decommission your old Direct Connect connection

Once you are confident the new Direct Connect connection meets your requirements, you can decommission one of the two old Direct Connect connections. In the following example, we have decommissioned the secondary Direct Connect connection (highlighted in figure 3 above in dotted purple color). After decommissioning of secondary Direct Connect connection, setup will look like as described in figure 4.

Direct Connect setup after decommissioning of secondary connection

Figure 4: Direct Connect setup after decommissioning of secondary connection

To decommission Direct Connect connection, first delete the VIFs and then delete the Direct Connect connection from your AWS Management Console. Work with your Direct Connect partner for the removal of physical circuits and the associated hardware.

Step 5: Swap primary and secondary connection

You may wish to use the new Direct Connect connection as the primary connection. To swap the primary and secondary connection, you can use the traffic engineering approach described in Step 2 using local preference community tags and local preference attribute to engineer the traffic flow as per your requirements. This is shown in the following diagram (figure 5).

Final setup after interchanging primary and secondary Direct Connect connections

Figure 5: Final setup after interchanging primary and secondary Direct Connect connections

Conclusion

As new Direct Connect locations open around the world, it’s possible that a new location is now available that will better serve your needs. Following the steps discussed in this post will help you plan and execute a smooth migration of a Direct Connect connection to a location closer to your on-premises network or branch location.

About the author

Ratan Kumar

Ratan Kumar is a Principal Solutions Architect at Amazon Web Services. A trusted technology advisor with over 20 years of experience working across a range of industry domains, Ratan’s passion lies in empowering enterprise customers innovate and transform their business by unlocking the potential of AWS cloud.