Q: We’re experiencing performance issues with Microsoft Teams in our Parallels Remote Application Server (RAS) environment. Users are complaining about poor audio/video quality and sluggish performance. What’s the best approach to optimize Teams for our on-premises RAS deployment? Scenario: We’re deploying Microsoft Teams across our Parallels RAS infrastructure. Half our session hosts have NVIDIA RTX A4000 cards, while the other half are CPU-only. How should we approach Teams optimization differently for each scenario, and what architecture considerations should we account for?
Architecture Foundation
When deploying Teams in a Parallels RAS environment, the key is understanding that you’re dealing with a multi-layered virtualization stack. Teams needs to efficiently traverse from the endpoint device through the RAS infrastructure to the hosted session, then back again—all while maintaining real-time performance for media workloads.
The presence of GPU hardware fundamentally changes how Teams processes media workloads in your RAS environment. Here’s what you need to architect for:
GPU-Enabled Hosts:
CPU-Only Hosts:
Scenario 1: GPU-Enabled RAS Hosts (NVIDIA RTX A4000)
Architecture Considerations
With GPU acceleration, your Teams deployment can handle significantly higher concurrent video sessions. Plan for:
Group Policy Configuration
GPU-Specific Teams Settings:
HKLM\SOFTWARE\Microsoft\Teams\IsVirtualDesktopOptimized = 1 (DWORD)
HKLM\SOFTWARE\Microsoft\Teams\DisableHWAcceleration = 0 (DWORD)
HKLM\SOFTWARE\Microsoft\Teams\DisableGpuAcceleration = 0 (DWORD)
RAS Multimedia Policies:
Resource Allocation
GPU Memory Management:
System Resources:
Scenario 2: CPU-Only RAS Hosts
Architecture Considerations
Without GPU acceleration, focus on CPU optimization and more conservative scaling:
Group Policy Configuration
CPU-Optimized Teams Settings:
HKLM\SOFTWARE\Microsoft\Teams\IsVirtualDesktopOptimized = 1 (DWORD)
HKLM\SOFTWARE\Microsoft\Teams\DisableHWAcceleration = 1 (DWORD)
HKLM\SOFTWARE\Microsoft\Teams\DisableGpuAcceleration = 1 (DWORD)
Performance Optimization Policies:
Resource Allocation
CPU Management:
Memory Allocation:
QoS Strategy: Universal Best Practices
Regardless of GPU presence, implement consistent QoS policies:
Traffic Classification:
Audio (DSCP 46):
├── Guaranteed: 100 kbps per session
├── Latency: <150ms
└── Loss tolerance: <0.1%
Video (DSCP 34):
├── GPU hosts: Up to 2.5 Mbps per session
├── CPU hosts: Up to 1.8 Mbps per session
├── Latency: <400ms
└── Burst allowance: 2x guaranteed rate
OS Optimization by Scenario
GPU Host Optimizations
NVIDIA-Specific:
Windows Configuration:
CPU Host Optimizations
Processor Configuration:
System Tuning:
Performance Benchmarks
GPU-Enabled Targets
CPU-Only Targets
Load Balancing Strategy
Intelligent Workload Distribution:
Monitoring KPIs by Configuration
GPU Host Monitoring
CPU Host Monitoring
Le bilan
Your dual-configuration approach requires thoughtful architecture but delivers significant benefits. GPU hosts can support 60-70% more concurrent video users while providing better quality. CPU hosts, when properly optimized, still deliver solid performance for mixed workloads.
Key Success Factors:
The investment in GPU acceleration typically pays for itself through improved user density and experience—just ensure your architecture takes full advantage of both configurations.
Got complex architecture scenarios? Let’s solve them together.
French 
English 
Hindi 