Mental Model¶

A framework for hypothesis-driven ACS diagnosis. Understanding the logical layers of communication helps isolate root causes.

Logical Layers of ACS¶

ACS interactions span from client devices to global cloud services. Isolating the layer of failure is the first step in troubleshooting.

graph TD
    Client[Client Device / SDK] --> Network[Network / Firewall / NAT]
    Network --> Service[Azure Communication Service]
    Service --> Carrier[Carrier / Mail Provider / Teams Service]
    Carrier --> Recipient[Recipient Device]

Layered Diagnosis Approach¶

1. SDK Layer (Client-side)¶

Symptoms: UI issues, initialization errors, token expiration, local device access failures.
Evidence: SDK logs, browser console errors, User Facing Diagnostics (UFD).

2. Network Layer¶

Symptoms: Timeouts, poor call quality, ICE negotiation failures, WebSocket disconnection.
Evidence: TURN/STUN accessibility, firewall logs, local bandwidth metrics.

3. Service Layer (Azure-side)¶

Symptoms: 429 throttling, 500 internal errors, 403 authorization issues.
Evidence: Azure Monitor Metrics, Log Analytics, Event Grid reports.

4. Configuration Layer¶

Symptoms: Verification failures, permission issues, resource access denied.
Evidence: Resource policy settings, domain verification status, token scopes.

5. Downstream Layer (Carrier/Recipient)¶

Symptoms: SMS not received, email bounced, Teams meeting access denied.
Evidence: Carrier delivery reports, SMTP bounce codes, Teams admin policy.

Communication Channel Model¶

Server-side Channels (SMS, Email)¶

Primarily asynchronous. Reliability depends on carrier handoff and domain reputation. Focus on delivery status reports.

Client-side Channels (Chat, Voice/Video)¶

Real-time and stateful. Reliability depends on persistent connections, network stability, and client device capabilities. Focus on latency and jitter.

Sources¶

Azure Communication Services Architecture Overview
Network Troubleshooting for Real-time Media