4.3 Branch Connector & Cloud Connector for Hybrid Clouds
Key Takeaways
- Zscaler Cloud Connector (CC) and Branch Connector (BC) are lightweight, containerized virtual appliances that automate secure overlay connectivity from public clouds (AWS, Azure, GCP) and enterprise branch offices to the Zero Trust Exchange.
- Cloud Connector delivers unified workload-to-internet (ZIA) security and workload-to-workload / workload-to-private-app (ZPA) connectivity directly from public cloud VPCs/VNets, eliminating the need for expensive Transit Gateways, complex BGP meshes, or virtual firewall clusters.
- Zero-Touch Provisioning (ZTP) leverages cloud-init userdata, infrastructure-as-code (Terraform/Ansible), and cloud-native template deployments to automatically bootstrap, license, and register connectors.
- Intelligent Traffic Forwarding dynamically classifies workload traffic at Layer 7, steering internet/SaaS flows to ZIA and private cross-cloud or cloud-to-on-prem flows to ZPA via inside-out DTLS/TLS tunnels.
- High Availability (HA) architectures integrate natively with cloud load balancers (AWS Gateway Load Balancer, Azure Standard Load Balancer, GCP Internal TCP LB) to achieve active-active scale and sub-second failover.
4.3 Branch Connector & Cloud Connector for Hybrid Clouds
Architectural Premise: Traditional multi-cloud networking relies on complex, fragile constructs: AWS Transit Gateways, Azure Virtual WANs, inter-cloud IPsec VPN meshes, centralized Next-Generation Firewall (NGFW) clusters, and intricate BGP route peering. Zscaler Cloud Connector (CC) and Branch Connector (BC) replace this legacy networking sprawl with zero-trust workload connectivity. Deployed as lightweight virtual appliances directly within VPCs, VNets, or branch virtualization clusters, they establish automated, inside-out micro-tunnels to the Zero Trust Exchange, securing Workload-to-Internet (ZIA) and Workload-to-Workload (ZPA) traffic without exposing routable subnets.
1. Cloud Connector & Branch Connector Overview
Cloud Connector and Branch Connector are purpose-built virtual appliances delivered as hardened Linux-based images (available as Amazon AMIs, Azure VHDs, GCP Images, and VMware ESXi/KVM OVAs). They act as intelligent edge proxies that combine the capabilities of a ZIA forwarding engine (utilizing Z-Tunnel 2.0 with DTLS/TLS encapsulation) and a ZPA App Connector/Client into a unified, automated software platform.
+-----------------------------------------------------------------------------+
| CLOUD & BRANCH CONNECTOR DEPLOYMENT SCOPE |
+-----------------------------------------------------------------------------+
Public Cloud Environments (AWS, Azure, GCP): On-Premises Branch & Virtualized Edge:
+-----------------------------------------+ +-----------------------------------+
| ZSCALER CLOUD CONNECTOR | | ZSCALER BRANCH CONNECTOR |
| * Deployed in VPC / VNet / GCP Project | | * Deployed on VMware ESXi / KVM |
| * Secures Cloud Workloads (EC2, VMs) | | * Secures Branch Servers & LANs |
| * Replaces Transit GWs & Cloud Firewalls| | * Replaces Complex Branch Routers |
+-----------------------------------------+ +-----------------------------------+
\ /
\ /
v v
+-----------------------------------------------------------------------------+
| ZSCALER ZERO TRUST EXCHANGE |
| [ ZIA: Workload-to-Internet ] [ ZPA: Workload-to-Workload ] |
+-----------------------------------------------------------------------------+
Key Distinctions: Cloud Connector vs. App Connector vs. Branch Connector
| Feature / Dimension | Zscaler Cloud Connector (CC) | Zscaler App Connector (AC) | Zscaler Branch Connector (BC) |
|---|---|---|---|
| Primary Function | Steers outbound workload traffic to ZIA & ZPA | Terminates inbound brokered connections to apps | Steers outbound branch LAN traffic to ZIA & ZPA |
| Traffic Types Handled | Workload-to-Internet (ZIA) & Workload-to-Workload (ZPA) | Private App Access (ZPA only) | Branch-to-Internet (ZIA) & Branch-to-Workload (ZPA) |
| Tunnel Protocol | Z-Tunnel 2.0 (DTLS/TLS over UDP/TCP 443) | Micro-tunnels to ZPA Service Edge (TLS) | Z-Tunnel 2.0 (DTLS/TLS over UDP/TCP 443) |
| Deployment Locations | AWS VPC, Azure VNet, GCP VPC, Nutanix | AWS, Azure, GCP, On-prem Data Center | Branch ESXi, Nutanix AHV, KVM, Linux Bare Metal |
| Inbound Ports Open | Zero (Inside-out outbound connections only) | Zero (Inside-out outbound connections only) | Zero (Inside-out outbound connections only) |
| Routing Integration | Cloud Route Tables, AWS GWLB, Azure Route Server | None (Proxies local TCP/UDP connections to server) | Local VLAN Gateway, PBR, BGP route peering |
2. Workload Traffic Steering: Workload-to-Internet & Workload-to-Workload
Cloud Connector provides intelligent, Layer 7-aware traffic forwarding by separating cloud outbound communications into two fundamental traffic streams:
+-----------------------------------------------------------------------------+
| DUAL-PATH WORKLOAD TRAFFIC FORWARDING |
+-----------------------------------------------------------------------------+
+------------------------------------------------+
| Cloud Workload (EC2 / Azure VM / Kubernetes) |
+-----------------------+------------------------+
|
Route Table Default (0.0.0.0/0) or Target Subnet
v
+------------------------------------------------+
| Zscaler Cloud Connector (CC) |
+-----------------------+------------------------+
/ \
Path A: Workload-to-Internet Path B: Workload-to-Workload (Private)
/ \
v v
+-----------------------------+ +-----------------------------+
| ZIA Public Service Edge | | ZPA Public/Private Edge |
| * SSL/TLS Deep Inspection | | * 1:1 Brokered Connection |
| * Advanced Threat Protect | | * Microsegmented App Access |
| * Data Loss Prevention (DLP)| | * No Transit Gateway Routing|
| * Cloud Sandbox Inspection | | * Overcoming Overlapping IPs|
+--------------+--------------+ +--------------+--------------+
| |
v v
[ Public Internet / SaaS ] [ Workload in Azure / DC ]
Path A: Workload-to-Internet (ZIA Forwarding)
When a cloud instance (e.g., an AWS EC2 instance updating Linux packages or accessing public SaaS APIs) sends traffic to an external destination:
- The cloud subnet route table directs
0.0.0.0/0traffic to the Cloud Connector instance interface. - Cloud Connector intercepts the connection, evaluates policy, and encapsulates the payload into a Z-Tunnel 2.0 (DTLS with TLS fallback) tunnel directed to the nearest ZIA Public Service Edge.
- ZIA applies the full enterprise security stack: Single-Scan Multi-Action (SSMA) engine, SSL Inspection, Cloud Firewall/IPS, DLP, and Cloud Sandbox.
- Clean traffic is forwarded to the public internet destination.
Path B: Workload-to-Workload / Workload-to-Private App (ZPA Forwarding)
When a workload in AWS needs to communicate with a database in Azure or a mainframe in an on-premises data center:
- The cloud workload initiates a connection to the destination FQDN or IP address (e.g.,
db.corp.azure.internal). - Cloud Connector intercepts the request, identifies it as a configured ZPA Application Segment, and acts as a ZPA client.
- It queries the ZPA Central Authority, establishes an authenticated, end-to-end encrypted micro-tunnel through the ZPA Service Edge, which stitches to the ZPA App Connector deployed adjacent to the database.
- No AWS Transit Gateway, Azure ExpressRoute, or Direct Connect peering is required, and the two clouds can utilize identical, overlapping RFC 1918 IP address spaces without NAT conflicts.
3. Zero-Touch Provisioning (ZTP) & Infrastructure as Code
Cloud Connector eliminates manual console provisioning by leveraging automated Zero-Touch Provisioning (ZTP) workflows.
+-----------------------------------------------------------------------------+
| ZERO-TOUCH PROVISIONING (ZTP) LIFECYCLE |
+-----------------------------------------------------------------------------+
1. ADMIN CONSOLE: Generate Provisioning URL & Template
[ ZIA / ZPA Admin Portal ] ---> Create "AWS-US-East-Connector-Group"
---> Generate Provisioning Token & Cloud-Init String
2. INFRASTRUCTURE AS CODE: Deploy via Terraform / CloudFormation
[ Terraform Script ] ---> Deploys VPC, Subnets, Route Tables
---> Launches Cloud Connector AMI with user_data = cloud-init
3. BOOTSTRAP & REGISTRATION: Automatic Cloud Enrollment
[ Cloud Connector VM Boots ] ---> Reads user_data provisioning token
---> Initiates Outbound TLS (TCP 443) to Zscaler Central Auth
---> Validates Token -> Downloads Security Configuration
---> Registers Health Status -> Begins Processing Traffic
Provisioning Parameters in Cloud-Init Userdata:
When launching a Cloud Connector instance in AWS or Azure, the orchestration script passes a YAML-formatted user_data payload containing:
zscaler_cc_provisioning_url: The unique cryptographic enrollment URL generated in the Zscaler portal.zscaler_cc_provisioning_token: The authentication token binding the virtual appliance to the enterprise tenant and Connector Group.zscaler_cc_group_name: The administrative group governing policy and location attributes.
4. High Availability (HA) and Cloud Integration Architectures
Enterprise cloud deployments require resilient, multi-AZ active-active high availability.
+-----------------------------------------------------------------------------+
| AWS GATEWAY LOAD BALANCER (GWLB) ACTIVE-ACTIVE HA |
+-----------------------------------------------------------------------------+
Workload Subnet (VPC)
+-------------------------------------------------------------------------+
| [ App Instance 1 (AZ-A) ] [ App Instance 2 (AZ-B) ] |
+--------------------+------------------------------------+---------------+
| |
v v
+-------------------------------------------------------------------------+
| AWS GATEWAY LOAD BALANCER (GWLB) |
| * Health checks CC instances via TCP/HTTP probes |
| * Encapsulates traffic using GENEVE protocol (UDP Port 6081) |
+--------------------+------------------------------------+---------------+
| (GENEVE) | (GENEVE)
v v
+-----------------------------+ +-----------------------------+
| Cloud Connector 1 (AZ-A) | | Cloud Connector 2 (AZ-B) |
| [ Active Appliance ] | | [ Active Appliance ] |
+--------------+--------------+ +--------------+--------------+
| |
+===================+====================+
|
v (Z-Tunnel 2.0 DTLS/TLS)
[ Zscaler Zero Trust Exchange ]
Cloud HA Design Patterns:
- AWS Integration: Cloud Connector integrates directly with AWS Gateway Load Balancer (GWLB) using GENEVE encapsulation (UDP 6081). Workload VPC route tables target the GWLB endpoint (
vpce-xxx), which load balances traffic across an autoscaling group of Cloud Connectors in different Availability Zones. - Azure Integration: Uses Azure Standard Internal Load Balancer (ILB) configured with HA Ports (
All Protocols / All Ports), distributing traffic across a Virtual Machine Scale Set (VMSS) of Cloud Connectors. Alternatively, Azure Route Server can peer via BGP with Cloud Connectors for dynamic route injection. - GCP Integration: Utilizes GCP Internal TCP/UDP Load Balancing with next-hop routing rules directing VPC egress traffic to backend Cloud Connector instances.
5. Connector Sizing and Form Factors
Zscaler Cloud and Branch Connectors are available in standardized sizing profiles tailored to throughput requirements:
| Sizing Profile | vCPU / RAM | Network Interfaces | Target Throughput | Typical Deployment Scenario |
|---|---|---|---|---|
| Small | 2 vCPU / 4 GB RAM | 2 (1 Mgmt, 1 Service) | Up to 200 Mbps | Small branch offices, dev VPCs, staging environments |
| Medium | 4 vCPU / 8 GB RAM | 2 (1 Mgmt, 1 Service) | Up to 500 Mbps | Medium branches, standard enterprise production VPCs |
| Large | 8 vCPU / 16 GB RAM | 2 to 4 interfaces | Up to 1 Gbps+ | Data centers, high-volume Kubernetes clusters, cloud hubs |
[!NOTE] Dual-Interface (Dual-Arm) Architecture: Cloud Connector utilizes two distinct network interfaces:
- Management Interface (
eth0): Used for appliance management, DNS resolution, SSH administration, and control plane communication with the Zscaler Central Authority.- Service Interface (
eth1): Dedicated dataplane interface that receives workload traffic, handles GENEVE/VLAN encapsulation, and establishes Z-Tunnel 2.0 dataplane tunnels to ZIA and ZPA.
6. Exam Traps & Real-World Pitfalls
[!WARNING] Exam Trap (Cloud Connector vs App Connector): Do not confuse the functional roles of Cloud Connector and App Connector. An App Connector is strictly an inbound listener that connects authenticated users/workloads to private applications hosted in that datacenter/VPC via ZPA. A Cloud Connector is an outbound gateway deployed to intercept traffic originating from workloads in that VPC and steer it outbound to ZIA (for internet) and ZPA (for private services).
[!IMPORTANT] Overlapping IP Resolution: When connecting multi-cloud VPCs (e.g., AWS VPC with
10.0.0.0/16and Azure VNet with10.0.0.0/16), legacy VPN/routing fails due to overlapping CIDRs. Because Cloud Connector and ZPA operate at Layer 7 by resolving application domain names to synthetic IPs and stitching broker micro-tunnels, cross-cloud communication succeeds seamlessly without complex NAT or re-IPing.
What is the primary operational distinction between a Zscaler Cloud Connector and a Zscaler App Connector?
An enterprise has deployed workloads in AWS VPC (10.0.0.0/16) and Azure VNet (10.0.0.0/16) with identical, overlapping IP address spaces. How does Zscaler Cloud Connector enable secure communication between these workloads without requiring complex Network Address Translation (NAT) or re-IPing?
How does Zero-Touch Provisioning (ZTP) automate the enrollment and deployment of Zscaler Cloud Connector instances in an AWS environment?
In an AWS high-availability architecture utilizing AWS Gateway Load Balancer (GWLB), which protocol is used to encapsulate workload traffic between the GWLB and the backend Cloud Connector instances?