多区域应用程序的 AWS 延迟改进

Question

We have clients from all over the world. So we are working to reduce latency.

Here is the schema that we currently use;

The problems are;

• All ec2s are linked to their own local RDS. We do it due to reducing the latency. But it very costly. Can we get the same latency as local rds using inter-region vpc with global RDS?

• All regions have their own Application Load Balancers. But it is too costly too. Is there a way to reduce application load balancer cost?

• Is there a better approach that you suggest to reduce latency with lower cost and better performance?

Thank you very much.

Answer 1

Can we get the same latency as local rds using inter-region vpc with global RDS?

Sadly, no. When your RDS and EC2 are in same region, the connections between them must cross, in the worst case scenario, from one AZ to other AZ. For inter-region connections you are increasing distance greatly, so latency will increase.

Regarding cost. RDS is very expensive. Maybe some of your applications don't need to use RDS? Self-managed db can be much cheaper run, but not easy to setup if you require advanced features such as fail-overs, read replicas, backups.

Also, if in a region you have multiple RDS instances, maybe you can consolidate them, and have one bigger RDS instead of several smaller?

Is there a way to reduce application load balancer cost?

If you have a typical setup of 1 ALB with 1 TG, then you can maybe reduce number of ALBs in a region by having 1 ALB for multiple TGs.

Answer 2

I don't exactly know what is the application, but you can have your infrastructure in one region and use an Edge-based API gateway in front of ALB to achieve lower latency across regions. The only downside is you might not have multi-region availability and only REST is supported, if it fits your need, you can experiment with this.