Application Load Balancer (ALB) - More than 1 ENI per availability zone

0

Hello,

Keeping the announced IPv4 cost in mind, starting the 1st of Februari 2024, I have a question regarding the amount of ENIs assigned to an ALB.

Currently, I am managing several accounts which contain multiple ALBs with ENIs deployed in all availability zones available to that specific region. Let's say that I have 3 availability zones to my disposal. This means that I should have 3 ENIs deployed in each availability zone for 1 ALB. After some initial analysis, I noticed that some ALBs contain more than 1 ENI per availability zone, which results in more than 3 ENIs per ALB. The total amount differs from ALB to ALB, sometimes I see 4 ENIs, sometimes 5 or 6...etc.

Could someone explain to me if I can predict the total amount of ENIs assigned to an ALB when deployed in all availability zones? To my knowledge it would deploy 1 ENI in each availability zone, but in reality, this seems not always to be the case. Could this be due to amount of applications connected to the ALB, which would cause a higher load? Due to our limited use of other types of load balancers, I cannot tell if this also occurs for these types. I already tried researching the AWS Docs, but I did not find (yet) an answer which I am looking for.

Thanks in advance!

Kind regards Tom K.

3 Answers
0
Accepted Answer

Could someone explain to me if I can predict the total amount of ENIs assigned to an ALB when deployed in all availability zones?

Define "predict" and "when deployed".
Short answer is No. ALB size for users is expressed in LCUs rather than individual nodes/ENIs. Corresponding CW metric is ConsumedLCUs. You can't say upfront whether N number of LCUs will always be equivalent to M number of ENIs. It may not even be 1 ENI per AZ when provisioned in case of pre-warming. I had a handful of such cases. https://repost.aws/questions/QUhvpfhsDsSZKRvoLOyWMRkQ/alb-pre-warming.

I recommend reaching out to AWS technical support if you have a question about specific ALB.

Val
answered 6 months ago
  • Thanks for the quick response. Most of the our shared load balancers contain several applications, where the LCU count is below 1 LCU (avg. of 0.6). This specific load balancer had 5 ENIs assigned which scaled-in back to 3 and I did not cross-reference this with the LCU count, which I will do in the future.

0

AWS offers "Public IP Insights" tool at no cost to track public IPv4 usage. This tool covers ALBs. Here is the reference to AWS doc: https://docs.aws.amazon.com/vpc/latest/ipam/view-public-ip-insights.html

Val
answered 6 months ago
  • Through the use of Public IP Insights, I came to the conclusion that some ALBs have more ENIs provisioned. Hence, this is not really an answer to my question. Maybe I should have made it more clear that I executed my analysis via this tool.

0

Those ENIs are created to allow ALB to talk to targets in your VPC. Minimum configuration is to have an ENI per AZ where ALB is deployed (2 or more) but when there is more traffic, ALB can scale more load balancer instances (not to be mixed with auto-scaling of target instances). As one ENI can be attached to single instances, more ENIs are created too.

Here is another question & answer about load balancer scaling that could be helpful

https://repost.aws/questions/QUdkDdF1yASQ65nlvaz3_WuA/about-dynamic-change-of-ip-address-when-using-elb

profile picture
EXPERT
Kallu
answered 6 months ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions