Elastic Load Balancer: why peak is so elevated and consumed so low?

The discrepancy between your peak and consumed Load Balancer Capacity Units (LCUs) for your Application Load Balancer (ALB) is quite common and can be explained by how these metrics are calculated and what they represent.

The "PeakLCUs" metric accounts for peaks in your traffic pattern across all scaling dimensions that the load balancer must support. This includes new connections, active connections, processed bytes, and rule evaluations. It represents the capacity your ALB needs to handle potential spikes in traffic, even if they are brief.

On the other hand, the "Consumed Load Balancer Capacity Units" (or ConsumedLCUs) metric represents the actual usage averaged over time and is used for billing purposes. This metric aggregates the billing dimensions of your traffic and is typically lower than the PeakLCUs, especially for low-traffic scenarios.

The reason your peak usage is showing 15-20 LCUs while your consumed LCUs are much lower is likely due to brief spikes in one or more of the scaling dimensions, even though your overall traffic is low. This could be caused by factors such as:

1. A high number of new connections in a short period
2. A spike in the number of active connections
3. A temporary increase in processed bytes
4. A large number of rule evaluations due to complex routing rules

As for billing, you are charged based on the ConsumedLCUs, not the PeakLCUs. The billing for your ALB consists of two main components:

1. An hourly charge for each hour or partial hour that your ALB is running
2. A charge based on the number of LCUs consumed

Given that your traffic is low, your actual bill for LCU usage is likely to be quite small, corresponding to the low "Consumed Load Balancer Capacity Units" shown in your graph.

If you want to optimize your costs further, you might consider reviewing your ALB configuration to ensure it's not over-provisioned for your current needs. However, keep in mind that the higher PeakLCUs provide capacity for handling potential traffic spikes, which can be beneficial for maintaining performance during unexpected increases in load.
Sources
ALB LCU capacity reservation metrics confusion | AWS re:Post
Why do I get billed for 4000+hours of alb usage for just two albs? | AWS re:Post
Using Load Balancer Capacity Unit Reservation to prepare for sharp increases in traffic | Networking & Content Delivery
Application and Network Traffic Distribution - Elastic Load Balancing Pricing - AWS