Network Synthetic Monitor


Fully managed, agentless synthetic monitoring for hybrid and cross-region AWS network paths — detect grey failures in minutes, not hours.

What is Network Synthetic Monitor? ▾

Amazon CloudWatch Network Synthetic Monitor is a fully managed feature of CloudWatch that provides continuous, synthetic network path monitoring. It sends active probe traffic from your VPC subnets to destinations — on-premises IPs (via Direct Connect or Site-to-Site VPN) or cross-region PrivateLink endpoints — and publishes real-time round-trip time (RTT) and packet loss metrics.

It requires no agents to install or maintain. You simply specify source subnets and destination IPs, and AWS provisions all necessary infrastructure (ENIs and probe instances) in the background.

Key capabilities: Monitors hybrid (Direct Connect, Site-to-Site VPN) and internal AWS paths (cross-region PrivateLink). Supports IPv4 and IPv6. Publishes a Network Health Indicator (NHI) for Direct Connect paths to pinpoint whether degradation is on the AWS side.

Problems It Addresses ▾

ProblemDescription
Grey FailuresIntermittent packet loss or fluctuating latency that the routing control plane (BGP) doesn't detect — but your applications suffer from reduced throughput and increased latency.
High MTTRWhen a network issue occurs, engineers spend significant time determining where the problem is — is it the AWS network, the Direct Connect partner backbone, or the on-premises network?
BGP Blind SpotsBGP uses TCP which tolerates degradation via retransmissions. A link can be dropping packets while BGP sessions remain up, leaving you unaware of the problem.
DIY Monitoring OverheadBefore Network Synthetic Monitor, teams had to build and maintain custom probe agents, EC2 instances running ping/traceroute scripts, or purchase third-party solutions.
Cross-Region Path BlindnessPrivateLink cross-region traffic traverses the AWS backbone, but without active probing there's no way to detect path-specific degradations between regions.
Multi-AZ Visibility GapsWorkloads in different Availability Zones may experience different network conditions. Traditional monitoring doesn't differentiate per-AZ network paths.

How Network Synthetic Monitor Solves These Issues ▾

SolutionDescription
Continuous active probingSends synthetic ICMP/TCP probes every 30–60 seconds from your subnets to your destinations — surfacing packet loss and latency spikes that BGP and passive monitoring miss.
Network Health Indicator (NHI)For Direct Connect paths, NHI uses anomaly detection to report whether degradation is within the AWS-controlled network (value = 1) or elsewhere (value = 0), dramatically reducing time to locate the fault.
Per-AZ, per-path granularityDeploy probes in each subnet/AZ to monitor specific network paths independently. Network engineers can correlate failures to individual DX connections or VPN tunnels.
Zero operational overheadAWS provisions and maintains all probe infrastructure (ENIs, instances). No agents to install, patch, or scale. ENI IP addresses persist even if a probe instance is replaced.
Native CloudWatch integrationMetrics flow directly into CloudWatch. Set static thresholds or use anomaly detection. Trigger automated failover via EventBridge + Lambda when degradation is detected.
Cross-region PrivateLink monitoringProbe PrivateLink interface endpoint IPs across regions using TCP, measuring the actual path your traffic takes across the AWS backbone.

Architecture Diagram ▾

The following diagram illustrates a network engineer's monitoring architecture using Network Synthetic Monitor with a Direct Connect High Resiliency setup. Probes are deployed in monitoring subnets across two Availability Zones, each targeting separate on-premises destinations reachable via distinct DX connections — enabling per-path fault isolation.

Network Synthetic Monitor architecture - Network Engineer monitoring setup with probes across AZs and Direct Connect paths
Figure: Network engineer monitoring architecture — probes in each AZ target specific DX connections for per-path visibility. Source: AWS Blog.

How It Works ▾

Step 1: Create a Monitor

Define a monitor specifying: aggregation interval (30 or 60 seconds), source VPC subnets, destination IPs, protocol (ICMP or TCP), port, and packet size.

Step 2: AWS Provisions Infrastructure

AWS creates managed probe instances and attaches customized ENIs to your specified subnets. Each source + destination + protocol + packet size combination is a "probe."

Step 3: Active Probes Transmit

ICMP: Sends echo requests, expects echo replies. TCP: Sends SYN packets, expects SYN+ACK. Periodically rotates source ports (1024–65535) for broader path coverage.

Step 4: Metrics Published to CloudWatch

Every aggregation interval, three metrics per probe are published under AWS/NetworkMonitor: RTT (μs), PacketLoss (%), and HealthIndicator (0 or 1).

Step 5: Alert & Automate

Configure CloudWatch Alarms (static thresholds or anomaly detection). Integrate with SNS for notifications or EventBridge + Lambda for automated failover actions.

Supported Protocols

ProtocolMechanismUse CaseNotes
ICMPEcho Request → Echo ReplyHybrid paths (DX, VPN)56 bytes default packet size
TCPSYN → SYN+ACKPrivateLink, service-port monitoringPort range 1024-65535 rotated for coverage. Required for PrivateLink (ICMP not forwarded).
Important: For PrivateLink monitoring, you must use TCP (not ICMP) since PrivateLink endpoints do not forward ICMP traffic. Set the port to match the service exposed by your NLB (e.g., 443, 80, or custom).

Monitoring Use Cases

Use CaseDescription
Hybrid / Direct ConnectMonitor on-premises connectivity from VPC subnets. Use NHI to distinguish AWS-side vs. customer-side issues.
Site-to-Site VPNDetect tunnel degradation that BGP sessions mask. Alert before application impact.
Cross-Region PrivateLinkProbe interface endpoint IPs across regions with TCP. Detect backbone latency or loss on your critical service paths.
Automated FailoverPair alarms with Lambda to automatically reroute traffic when a path becomes impaired.

Key Metrics ▾

Network Synthetic Monitor publishes the following metrics under the AWS/NetworkMonitor namespace at configurable intervals:

Aggregation
30s
Minimum interval
Protocols
ICMP / TCP
Configurable per probe
Setup
Agentless
Fully managed by AWS
  • RTT (μs) — Round-trip time in microseconds
  • PacketLoss (%) — Percentage of probes that did not receive a response
  • HealthIndicator (0 or 1) — For Direct Connect paths, indicates whether degradation is on the AWS side (1 = AWS issue detected)