citrix-infrastructure-design（思杰基础设施设计）

Citrix Infrastructure Design

Overview

This skill provides guidance for designing and architecting Citrix infrastructure, including component sizing, high availability patterns, multi-site deployments, and disaster recovery planning.

Architecture Patterns

Single-Site Architecture

Components:

• 2+ Delivery Controllers (N+1 redundancy)
• SQL Server (AlwaysOn or mirroring)
• StoreFront server group (2-5 servers)
• NetScaler ADC HA pair
• Hypervisor cluster

Use Cases:

• Small to medium deployments
• Single geographic location
• Simpler management requirements

Multi-Site Architecture

Components per site:

• Local Delivery Controllers
• Local StoreFront servers
• Local database replica
• Zone configuration in single site

Cross-site components:

• GSLB for user routing
• Database replication
• Image management strategy

Use Cases:

• Geographic distribution
• DR requirements
• Regional performance needs

Citrix Cloud Hybrid

Citrix Cloud (managed):

• Delivery Controllers
• Site database
• Studio/Director
• Licensing
• Gateway Service

Customer managed:

• Cloud Connectors (2+ per location)
• VDAs and applications
• Hypervisor/cloud infrastructure
• Active Directory

Benefits:

• Reduced infrastructure
• Automatic updates
• Built-in redundancy

Component Sizing

Delivery Controllers

Placement:

• Anti-affinity rules across hosts
• Different failure domains
• Separate from other workloads

SQL Database

Database sizing:

• Site DB: 500 MB - 2 GB
• Logging DB: Grows based on retention
• Monitoring DB: Grows based on retention

VDA Sizing

Single-Session (VDI):

Multi-Session (RDSH):

Storage Considerations

MCS Requirements:

• Identity disk: 16 MB per VM (fixed)
• Difference disk: 10-40 GB (grows with writes)
• Write cache: 10-20 GB SSD/NVMe recommended

IOPS Guidelines:

• Boot storm: 50-100 IOPS per VM (brief)
• Steady state: 5-30 IOPS per VM
• Use MCS I/O with RAM cache to reduce storage load

High Availability Design

Controller Redundancy

# Local Host Cache provides outage protection
# Minimum 2 controllers per site
# Controllers share site database
# Automatic failover to LHC during outage

LHC Capabilities:

• VDA registration
• Session brokering
• Power management
• Some policy application

LHC Limitations:

• No new configurations
• No monitoring data
• Limited reporting

Database HA

SQL AlwaysOn Availability Groups (Recommended):

• Synchronous replication within site
• Automatic failover
• Multiple readable secondaries

SQL Mirroring:

• Simpler setup
• Single secondary
• Manual or automatic failover

Connection String:

Server=SQLListener.domain.com;Database=CitrixSite;
MultiSubnetFailover=True;Integrated Security=True

StoreFront HA

• Server group with 2-5 members
• Sub-40ms latency required
• Configuration auto-synchronized
• Load balanced via NetScaler or NLB

NetScaler HA

• Active/Passive or Active/Active
• Synchronous configuration
• Virtual MAC for seamless failover
• Health monitoring for backend services

Network Design

VLAN Segmentation

Firewall Requirements

User to StoreFront:

• 443/TCP (HTTPS)

StoreFront to Controller:

• 80/TCP (HTTP)
• 443/TCP (HTTPS)

User to VDA (internal):

• 1494/TCP (ICA)
• 2598/TCP (Session Reliability)
• 443/TCP (TLS)

VDA to Controller:

• 80/TCP (Registration)
• 443/TCP (Secure registration)

Bandwidth Planning

Disaster Recovery

RTO/RPO Objectives

Active/Active Design

• Controllers at both sites
• GSLB routes users to nearest site
• Real-time database replication
• Both sites handle production load
• Automatic failover

Active/Passive Design

• Secondary site in standby
• Asynchronous replication acceptable
• Manual or automated failover
• Lower cost than active/active
• Longer RTO

DR Considerations

Non-persistent VDI advantages:

• No VM replication needed
• Rapid provisioning at DR site
• Only master images replicated
• Lower storage costs

Key data to replicate:

• Site database
• Master images
• User profiles (if persistent)
• Application data

Design Best Practices

Scalability

1. Design for growth - Size for 3-5 year projection
2. Use zones - Separate workloads by zone
3. Catalog organization - Group by OS, workload, location
4. Delivery group strategy - Align with user groups

Performance

1. MCS I/O - Use RAM cache with disk overflow
2. Storage tiering - SSD for write cache, HDD for read
3. Network optimization - EDT, Multi-Stream ICA
4. Profile optimization - Streaming, exclusions

Security

1. Segmentation - Isolate management plane
2. Encryption - TLS for all communications
3. MFA - Required for external access
4. Least privilege - RBAC for administration

Operations

1. Monitoring - Director, SCOM, third-party
2. Alerting - Proactive threshold alerts
3. Automation - PowerShell for routine tasks
4. Documentation - Architecture, runbooks, DR

Reference Materials

For detailed design guidance, see:

• citrix-knowledge/domain-knowledge/comprehensive-citrix-knowledge.md
• citrix-knowledge/architecture/ for design patterns
• citrix-knowledge/templates/ for documentation templates