Skip to main content
Skip table of contents

AMD Cloud Design Standards

Overview

https://youtu.be/qGtjFBwVkmE

Cloudamize and AMD have collaborated on three distinct designs that allow customers to target instance families backed by AMD CPUs. These designs incorporate the following core principles:

  • Right-Sized Recommendations: Instances are selected based on actual performance telemetry from on-premises workloads.

  • License Optimization: Leverages Reduced Core CPU Optimization (setting specific vCPU counts) to minimize Windows and SQL Server licensing overhead while maintaining performance.

  • Economic Assumptions: All designs assume AWS On-Demand pricing and a 10% Network Load Factor (assuming 90% of traffic remains within the AWS network).

Design AMD Lowest Cost

Designed for cost-sensitive environments and workloads with variable utilization (e.g., Dev/Test).

  • Processor Generations: Newest (5th Gen), Previous (4th Gen), and Legacy (3rd Gen) AMD EPYC.

  • Allowed Instance Families:

    • Burstable: T3a (Included)

    • Compute: C6a, C7a, C8a

    • General: M6a, M7a, M8a

    • Memory: R6a, R7a, R8a, X8aedz

  • Service Level Targets (SLT):

    • Target Peak CPU Threshold: 90%

    • Storage Capacity: 100%

Design AMD Easy Path

Designed for stable, production-ready migrations requiring a predictable 1-to-1 hardware mapping.

  • Processor Generation: Only 3rd Gen AMD EPYC "Milan".

  • Allowed Instance Families:

    • C6a (Compute), M6a (General), R6a (Memory).

  • Exclude burstable (T) instance types to ensure dedicated CPU resources.

  • Mapping: Fixed hardware mapping to ensure on-premises parity.

Design AMD Highest Performance

Designed for mission-critical, high-throughput applications (e.g., HPC, Big Data, SQL Enterprise).

  • Primary Generation: Newest Generation Only (5th Gen AMD EPYC "Turin").

  • Primary Instance Families:

    • C8a, M8a, R8a, X8aedz

  • Fallback Logic: If the newest generation is unavailable in a region, use:

    1. C7a, M7a, R7a

    2. C6a, M6a, R6a

  • Constraints: Exclude burstable (T) instance types.

  • Service Level Targets (SLT):

    • Target Peak CPU Threshold: 70% (Ensures 30% headroom for peak performance).

    • Storage Capacity: 120% (Over-provisioned for high-speed I/O).

If you have any further queries, please reach out to the Cloudamize Helpdesk at helpdesk@cloudamize.com.

JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.