Routing Preferences for Service Connection Traffic

How Prisma Access routes its service connection traffic from mobile users and remote network connections, and the different modes you can use.
Prisma Access uses BGP for dynamic routing, and uses BGP path selection to install routes in the route table. When Prisma Access routes traffic to your headquarters or data center using service connections, it uses routing methods that direct that traffic effectively. Prisma Access uses a default routing model that was designed to fit the majority of network deployments; however, not all organization’s networks are the same. To fit a wider range of deployments, Prisma Access allows you choose another mode for service connection routing. The following sections describe the BGP routing methods that Prisma Access uses, along with the factors you need to consider in your organization’s network before changing Prisma Access’ default method of service connection routing.
Changing the Prisma Access service connection routing method requires a thorough understanding of your organization’s topology and routing devices, along with an understanding of how Prisma Access routing works as described in this section. We recommend that you read this section carefully before changing the routing method from the default setting.
Prisma Access supports static routing and dynamic routing using BGP for service and remote network connections; this section assumes that you use BGP routing for your Prisma Access deployments. When you select BGP routing, your organization’s network learns BGP information from Prisma Access.

Routing Modes for Service Connections

You can choose from the following routing modes with Prisma Access:
  • Default routing—This is the current routing model that Prisma Access uses.
    Use this routing mode if you want Prisma Access to use BGP best path-selection mechanisms without adjusting any of the BGP attributes. In this mode, Prisma Access will honor any attribute advertised by the customer premises equipment (CPE).
  • Hot Potato Routing—Prisma Access hands off the traffic as quickly as it can to your organization’s network.
    Use this routing method if you want your organization’s network to perform the majority of routing decisions.

Mobile User and Remote Network Routing to Service Connections Overview

It is useful to understand how Prisma Access routes traffic between mobile users, remote networks, and service connections, because the routing used by mobile user traffic and remote network traffic between service connections is different.
Mobile User-service connection routing
—The mobile user connection forms an IPSec tunnel with the nearest service connection. Prisma Access uses iBGP for internal routing and eBGP to peer with the customer premises equipment at the data center. The following diagram shows mobile users in Regions 1 and 2 being routed to the respective service connections in that region. Mobile users in Region 1 are accessing applications
located at Data Center 1. If your organization’s network uses BGP routing for their service connections and a service connection experiences an ISP failure at Data Center 1, Prisma Access detects the failure and routes the traffic for applications
to Data Center 2 after BGP convergence, providing redundancy to your network’s data centers.
Prisma Access uses the following timing with BGP when it detects a failure: If you configure BGP routing and have enabled tunnel monitoring, the shortest default hold time to determine that a security parameter index (SPI) is failing is the tunnel monitor, which removes all routes to a peer when it detects a tunnel failure for 15 consecutive seconds. In this way, the tunnel monitor determines the behavior of the BGP routes. If you do not configure tunnel monitoring, the hold timer determines the amount of time that the tunnel is down before removing the route. Prisma Access uses the default BGP HoldTime value of 90 seconds as defined by RFC 4271, which is the maximum wait time before Prisma Access removes a route for an inactive SPI. If the peer BGP device has a shorter configured hold time, the BGP hold timer uses the lower value. When the secondary tunnel is successfully installed, the secondary route takes precedence until the primary tunnel comes back up. If the primary and secondary are both up, the primary route takes priority.
Remote Network-service connection routing
—Prisma Access creates a full mesh network with other remote networks and service connections. As with mobile users, Prisma Access uses iBGP for its internal routing and eBGP to peer with customer premises equipment to exchange routes. If a user in Branch 1 is accessing application
from Data Center 1 in your organization’s data center and the link between Branch 1 and Data Center 1 goes down, Prisma Access routes the traffic for application
to Data Center 2 after BGP convergence.

Prisma Access Default Routing

The following figure shows an example of Prisma Access routing service connection traffic in default routing mode. The organization’s network has three separate networks in three data centers and does not have a backbone connecting the networks. In default routing mode, mobile user pools are advertised equally on the three networks, as shown at the bottom of the figure.
Note that, when Prisma Access advertises mobile user routes, it divides the subnets into Class C /24 address blocks before advertising them; thus, it advertises the /20 mobile user subnets in chunks of /24 as prefixes are consumed by the gateways.
Make a note of how Prisma Access uses BGP route advertisements:
  • Prisma Access does not adjust the default BGP attributes for mobile user advertised routes (Prisma Access adds its AS number to the route advertisements).
  • Prisma Access advertises mobile user routes in blocks of /24 subnets and adds BGP community values in the routes it advertises through the service connection. The following figure shows a mobile user deployment with three service connections and three different IP address blocks specified for the mobile user IP address pool:
    for the
    Asia, Australia & Japan
    for the
    Africa, Europe & Middle East
    region, and
    for the
    North America & South America
    region. Prisma Access divides these routes into block of /24 and advertises them with an Prisma Access’ AS number of
    , but also appends the BGP community values to the advertisements (
    for Asia,
    for EU, and
    for US). Those routes are shown in the middle of the figure. In this way, you can differentiate service connections in your network, even though Prisma Access assigns the same AS number to them.
You can view the community string by selecting
Cloud Services
Network Details
Service Connection
Show BGP Status
and find the
field in the
The following figure shows a more common network with a full-mesh eBGP backbone. The figure shows the routes that Prisma Access has learned from your organization’s network on the top right. Note the extra routes that Prisma Access has learned through the Prisma Access backbone (iBGP) and your organization’s backbone (eBGP).
For traffic between mobile users in the
North America & South America
region (US in the diagram) and the data center in your organization’s
Africa, Europe & Middle East
region (EU in the diagram), Prisma Access chooses the path through the EU service connection because it prefers routes with a shorter AS-PATH.
In deployments with a full-mesh eBGP backbone, asymmetry can arise when Prisma Access cannot reach a particular data center due to an ISP/CPE failure at the customer’s data center. The following figure shows what could happen when the link to the EU service connection goes down. Your network detects the link failure and builds a new route table for AS 200. Traffic from the US service connection to AS 200 uses the path through AS 100 because the eBGP route for your backbone between AS 200 and AS 100 is preferred to the iBGP route between service connections EU and US. However, return traffic is not guaranteed through the same path because the on-premise CPE can choose either path (shown in red) to return the traffic.
The previous examples show a network whose routes have not been aggregated (that is, you have not performed route summarization before you send the BGP route advertisements to Prisma Access). The following example shows a network that summarizes its routes to before sending to Prisma Access. If you select default routing, this configuration can lead to asymmetric routing issues, because Prisma Access cannot determine the correct return path from the summarized routes.
If your Prisma Access deployment has Remote Networks, Palo Alto Networks does not recommend the use of route summarization on Service Connections. Route summarization on service connections is for Mobile Users deployments only.
If you use route aggregation for mobile users, we strongly recommend that you enable hot potato routing instead of default routing, where Prisma Access hands off the traffic as quickly as possible to your organization’s network; in addition, we recommend that you select a
Backup SC
as described in the following section for each service connection to have a deterministic routing behavior.

Hot Potato Routing

When you select
Hot Potato Routing
, Prisma Access egresses the traffic bound to service connections/data centers from its internal network as quickly as possible.
With hot potato routing, Prisma Access prepends the AS path (AS-PATH) to the BGP prefix advertisements sent from gateways. This prepending is performed when the prefixes are advertised out of the service connection to your organization’s on-premise CPE. Prisma Access prepends the AS-PATHs so that your CPE gives the correct preference to the primary and secondary tunnels, so that if the primary tunnel goes down, your CPE chooses the secondary tunnel as the backup.
If you specified a different IP address for the secondary (backup) BGP peer, Prisma Access adds more prepends based on the tunnel type, as shown in the following table.
Prefix Type
Service Connection Tunnel Type
Number of As-Path Prepends
Total AS-PATHs Seen on the CPE
Gateway prefixes from primary service connection
Primary or Secondary tunnel with the same BGP peer IP address
Gateway prefixes from backup service connection
Primary or Secondary tunnel with the same BGP peer IP address
Gateway prefixes from all other service connections
Primary or Secondary tunnel with the same BGP peer IP address
Gateway prefixes from primary service connection
Secondary tunnel with a different BGP peer IP address
Gateway prefixes from backup service connection
Secondary tunnel with a different BGP peer IP address
Gateway prefixes from all other service connections
Secondary tunnel with a different BGP peer IP address
In hot potato routing mode, Prisma Access allows you to specify a backup service connection (Backup SC) during onboarding. Specifying a
Backup SC
informs Prisma Access to use that service connection as the backup when a service connection link fails.
The following figure shows a hot potato routing configuration for traffic between the US service connection and AS 200, with the EU service connection configured as the
Backup SC
of the US connection. Using hot potato routing, Prisma Access sends the traffic from its closest exit path through the US service connection. The return traffic takes the same path through AS100 because this path has a shorter AS-PATH to the mobile user pool in the US location. Prisma Access prepends the AS-PATH to its prefix advertisements depending on whether the tunnel is a primary tunnel, a backup tunnel, or not used for either primary or backup.
Because you have set up a backup service connection, if the link to the US service connection goes down, hot potato routing sends the traffic out using its shortest route through the EU service connection. This routing scenario also applies to networks that use route aggregation.
You can also use backup service connections for multiple service connections in a single region. The following figure shows a Prisma Access deployment with two service connections in the North America region. In this case, you specify a
Backup SC
of US-E for the US-W service connection, and vice versa, to ensure symmetric routing.

Configure Routing Preferences

To enable routing preferences, complete the following steps.

Recommended For You