Right-sizing of Nodes in the Cloud
Right-sizing of nodes in the cloud is important to save costs and resources (energy & embodied carbon).
CPU- vs. Memory-Bound Instances
One factor, that leads to the underutilization of servers (and the difficulty for users to pick smaller instances) is a high usage of memory from the services hosted inside VMs. If VMs were more CPU-bounded, it would be easier for cloud providers to use oversubscription mechanisms to ensure a high usage of their servers. Modern software stack tends to use plenty of memory. In fact, over-using memory (through caching and memorization) is often seen as a "green coding" practice even if high memory usage in VMs tends to produce under-used servers.
β LinkedIn-Comment by Dimitri Saingre
AWS:
π References
better: fewer nodes (bigger instances)
Smaller doesnβt always mean less. While the flexibility and initial budget-friendliness of using smaller nodes are tempting, they easily lead to higher energy use, lower utilisation, and spiraling inefficiencies.
While this is not a call to just go for the biggest machines available to us, for now, the quickest way for earth-conscious IT practitioners to help reduce carbon emissions is to revisit our node size calculations in the past and see if we might be better off altering our machines to better reflect the value that we are creating.
AWS Cost Management β Right Sizing
Right sizing is the process of matching instance types and sizes to your workload performance and capacity requirements at the lowest possible cost. Itβs also the process of looking at deployed instances and identifying opportunities to eliminate or downsize without compromising capacity or other requirements, which results in lower costs.