Optimization Solver is professional software for solving optimization problems across fields including electrical energy, industrial manufacturing, transportation and logistics, retail, finance, and cloud computing. It is the core of industrial design software and helps enterprises design or optimize production solutions, allocate resources efficiently, and improve decision-making.
Overview
Optimization Solver is developed by the MindOpt Solver team at the Decision Intelligence Lab of Alibaba DAMO Academy. It is the core of industrial design software and solves optimization problems across industries — from elastic cloud resource scheduling to energy and manufacturing. It saves Alibaba Cloud hundreds of millions of US dollars annually in elastic computing resource scheduling and optimization. The software helps design or optimize production solutions, appropriately allocate resources, and improve decision-making in various scenarios.
This figure, provided by the MindOpt Solver team, illustrates the problem types, technical capabilities, and business benefits of Optimization Solver.
Features
Optimization Solver supports three solving paradigms, each designed for a different class of optimization problem:
Mathematical programming solving — for problems fully defined by objective functions, variables, and constraints
Simulation optimization — for complex problems that cannot be expressed analytically or whose constraints are not quantifiable
-
Online optimization — for real-world systems running under uncertainty
Mathematical programming solving
Mathematical programming addresses problems defined by a quantifiable formula — an objective function, decision variables, and constraints. It covers linear programming (LP), nonlinear programming (NLP), and mixed-integer programming (MIP).
The following problem types are currently supported:
Linear programming (LP)
Convex quadratic programming (QP)
Semidefinite programming (SDP)
Mixed-integer linear programming (MILP)
Support for additional problem types is under active development.
Simulation optimization
Simulation optimization handles problems that resist analytical formulation — where objective functions are unavailable, constraints are not fully quantifiable, or the system itself is a black box. It includes black-box optimization and zeroth-order (ZO) optimization.
Optimization Solver queries a simulation system with control parameters (input variables) and uses the returned evaluation results to search for an optimal solution, without requiring an explicit model of the system.
Typical use cases:
Policy search in reinforcement learning
Solution design for industrial smelting processes
Budget quota optimization for computing resources
Online optimization
Online optimization applies to live systems operating under incomplete information. Rather than solving a problem offline and applying results later, it optimizes decisions in real time as the system runs.
It is commonly used in online platforms — including e-commerce, video streaming, and advertising — for the following scenarios:
Product selection and material curation
New product recommendations
Traffic throttling
Real-time distribution of user rights and interests