AlphaQUBO is a solver for optimization problems and uses QUBO (Quadratic Unconstrained Binary Optimization) formulations as input.

AlphaQUBO can optimize problems in a number of application areas, including:
o Resource allocation
o Molecule discovery
o Portfolio optimization
o Supply chain optimization
o CyberSecurity ( anomaly detection )

Meta-Analytics suggests the memory optimized series of EC2 instance types for use in the ECS cluster. While multiple QUBO problems can be solved in parallel, each QUBO problem is solved on one EC2 instance. The primary resource consumed by AlphaQUBO is memory which in turn is highly dependent on each QUBO solved. Below are suggested starting points for EC2 instance and memory requirements.

To choose the instance type to execute on, please use the following RAM requirements. We've listed the recommendations from the AWS Price Calculator for the most cost effective instance type for x86_64. We recommend you start with these instance types and if you use case has more stringent performance needs, choose an instance type focused on memory optimization. As you choose different instance types, please adjust the CPU and memory constraints needed for a task.

Please note, each QUBO problem may behave differently based on the individual problem.

10,000 variables : r5a.large (10 GiB RAM)
100,000 variables : r5a.2xlarge (40 GiB RAM)
200,000 variables : r5a.8xlarge (150 GiB RAM)
300,000 variables : r5a.16xlarge (500 GiB RAM)
400,000 variables : x1.16xlarge (940 GiB RAM)