Selecting Appropriate Compute Resources

Evaluating compute demands involves the specific compute demands of workloads, including factors such as instance types, scalability, and containerization. Different compute services have different features and characteristics that can impact workload performance. Select the most appropriate compute service to ensure workloads run efficiently. Consider the following strategies:

• Instance types

Different instance types are optimized for different workloads, such as CPU-optimized, memory-optimized, and GPU-optimized. Choose the instance type that meets your needs.

• Auto scaling

If workload demands are unpredictable, consider compute services with auto scaling capabilities that can automatically adjust compute capacity as required. Auto scaling ensures sufficient resources during peak hours and prevents over-allocation during off-peak hours.

• Containerization

Containers have performance advantages over non-containerized workloads. If containers fit your architecture requirements, consider containerization. Containers improve compute performance through isolation, resource efficiency, fast startup times, and portability.

When using containers, consider design factors such as containerizing all application components. Linux-based container runtimes are used for lightweight images to provide a short lifecycle for containers so that they are immutable and replaceable. Relevant logs and metrics are collected from containers, container hosts, and basic clusters to monitor and analyze performance. Containers are just one component of the overall architecture. An appropriate container business process coordinator (such as Kubernetes) can enhance performance and scalability.

• PERF03-01 Selecting Appropriate Compute Services
  
• PERF03-02 Selecting VMs and Container Nodes of Appropriate Specifications
  
• PERF03-03 Applying Auto Scaling