Coconote
AI notes
AI voice & video notes
Try for free
🌐
Azure Proximity Placement Groups - Key Points and Best Practices
Jun 27, 2024
Proximity Placement Groups
Introduction
Proximity Placement Groups (PPGs) in Azure
Key considerations: best practices for usage
Target audience: users focusing on minimizing latency between services and VM scale sets
Separating Resources in Azure
Separation for resiliency and isolation from different problem levels:
Node level, Rack level, Data Center level, Regional level
Mechanisms for separation:
Availability Sets:
Distributes workloads among different fault domains/racks
Availability Zones:
Group of buildings within a region, provides isolation from building-level problems
Regions:
Separate physical locations with considerable distances, useful for disaster recovery
Proximity and Separation Effects
Proximity as a side effect of separation techniques:
Availability Sets give some proximity within clusters
Regions offer a ~2ms round trip latency envelope
No SLA guaranteeing proximity
Proximity Placement Groups (PPGs)
Purpose:
Minimize latency between VMs and VM scale sets
Applicability:
Only for VMs and VM scale sets, not other services
Expectation:
Sub-millisecond latency (not guaranteed)
Working Mechanism of PPGs
Creation and Pinning:
Create a PPG in a chosen region (e.g., South Central US PPG1)
PPG is initially floating in Azure Resource Manager, not pinned to a data center
First resource added to PPG pins it to a specific data center
All subsequent resources in that PPG will be in the same data center
Availability and Capacity Constraint:
Capacity limited to the pinned data center
Supported VM SKUs depend on the specific data center
Not About Disks:
PPGs do not reduce latency between VMs and their disks
Best Practices for Using PPGs
Declarative Technologies:
Use templates (ARM JSON, Bicep, Terraform) for VMs and PPGs
Helps Azure Resource Manager make informed placement decisions
Non-Declarative Approaches:
Avoid using Powershell, CLI, or the portal for initial PPG setup
If necessary, start with the most exotic VM SKU to improve placement success
Handling Constraints and Errors
Over-Constrained Allocation Request:
Indicates unmatched SKU combination in a single data center
Retry by reordering resources or use a declarative template
Additional Considerations
Existing VMs:
Can de-allocate, add to PPG, and start up
Maintains alignment in the PPG
Availability Sets in AZs via PPGs:
Allows placing availability sets in a specific Availability Zone
Maintenance and Re-alignment:
Resources may show as non-aligned after maintenance
De-allocate and restart to re-align
Conclusion
Purpose:
PPGs aim to reduce VM-to-VM latency
Not Influence on Disks:
Does not influence VM-to-disk latency
Best Practice:
Use accelerated networking to further reduce latencies
📄
Full transcript