Which type of scaling can eliminate downtime during compute instance adjustments?

Horizontal scaling is the correct choice because it involves adding or removing compute instances to manage workload demands without impacting the availability of applications. This approach allows resources to be adjusted seamlessly; new instances can be added to handle increased loads, and existing instances can be removed when demand decreases, all while maintaining service continuity.

Horizontal scaling is preferred in cloud environments, particularly in microservices architectures, as it enhances resilience and can optimize resource use efficiently. This elasticity is beneficial for applications that need to remain available during peak and off-peak times, mitigating the potential for downtime that can occur with other scaling strategies.

In contrast, vertical scaling involves resizing a single instance by adding more resources (like CPU or memory), which usually requires downtime as the instance often needs to be stopped and started. Dynamic scaling can automate the scaling process based on pre-defined metrics, but whether it involves horizontal or vertical scaling can lead to downtime if not managed properly. Manual scaling requires human intervention to adjust resources, which can also lead to potential service interruptions if not executed swiftly.