Lab Objective
Define an SLA class specifying acceptable loss, latency, and jitter thresholds, configure a centralized data policy matching voice traffic and applying that SLA class with a preferred transport color, and verify voice traffic is steered onto the compliant path while other traffic uses the default path.
Lab Purpose
Policy-Based Routing, covered earlier in this series, required configuring a route-map on each individual router to steer specific traffic. SD-WAN's centralized data policy achieves a conceptually similar outcome — steering traffic based on criteria beyond simple destination-based routing — but is authored once at the controller and pushed to every relevant edge device automatically.
Lab Topology
Edge1 (Site 100) has two transports:
biz-internet color (higher latency)
mpls color (lower latency, preferred
for voice)
Edge2 (Site 200) reachable via both colors
Goal: voice traffic (UDP 16384-32767) should
prefer the mpls transport when it meets SLATask 1: Define an SLA Class
Create an SLA class specifying maximum acceptable loss, latency, and jitter for voice-quality traffic.
Task 2: Create a Centralized Data Policy Matching Voice Traffic
Define a data policy matching the UDP port range typically used for RTP voice traffic.
Task 3: Apply the SLA Class with a Preferred Color
Configure the policy action to apply the SLA class, preferring the mpls color when SLA is met.
Task 4: Apply the Policy to the Relevant Site
Apply the centralized policy to Site 100's edge devices.
Task 5: Verify Voice Traffic Follows the Preferred Path
Generate voice-pattern traffic and confirm it uses the mpls transport, while other traffic continues using the default path selection.
Solution and Verification
-- Centralized policy configuration
-- (conceptually entered at the controller,
-- shown here in policy CLI form):
policy
sla-class VOICE-SLA
loss 1
latency 150
jitter 30
data-policy VOICE-STEERING
vpn-list VPN-1
sequence 10
match
source-data-prefix-list ANY
dscp 46
action accept
set-service sla-class VOICE-SLA preferred-color mplspolicy
apply-policy site-list SITE-100
data-policy VOICE-STEERING from-serviceEdge1# show sdwan policy from-vsmart
...
data-policy VOICE-STEERING
vpn-list VPN-1
sequence 10
match: dscp 46
action: accept, sla-class VOICE-SLA,
preferred-color mpls
-- Confirms the centrally authored policy
-- has been pushed down and is active on
-- this edge, without any local CLI policy
-- configuration having been typed on Edge1
-- itselfIPPhone1> [places a call, generating RTP
traffic marked DSCP EF/46]
Edge1# show sdwan app-route stats
TLOC-COLOR LOSS LATENCY JITTER SLA-CLASS
mpls 0 28ms 3ms VOICE-SLA (met)
biz-internet 0 95ms 18ms VOICE-SLA (met)
-- Both paths currently meet the SLA, but
-- mpls is the configured preferred-color,
-- so voice traffic is steered there
-- specificallyEdge1# show sdwan policy app-route-stats | include mpls
...voice flow matched, using tloc mpls...
-- Confirms voice-marked traffic is
-- following the mpls path as intended
OtherApp-PC> [generates non-voice traffic]
Edge1# show ... (non-voice flow)
...default path selection applied, not
subject to VOICE-STEERING policy...
-- Non-matching traffic continues using
-- ordinary path selection, unaffected by
-- the voice-specific policyKey Takeaway
SD-WAN's centralized data policy is authored once and automatically distributed to every applicable edge, in contrast to the earlier PBR lab where the identical route-map logic would need to be manually replicated on every router requiring the same steering behavior — this centralization is precisely what makes consistent application-aware routing practical across dozens or hundreds of sites, since a single policy change at the controller propagates everywhere it applies without touching individual device configurations.