Rebalancer
An open-source solver from Meta for data center task placement and optimization, built with a C++ core and Python bindings. It supports constraint modeling, multi-objective priorities and multiple solving backends.
Timeline (last 90 days)
Meta's open-source Rebalancer is a solver for data center task placement and optimization, providing a C++ core and Python bindings.
Rebalancer's core idea is to assign tasks to racks and data centers while satisfying various constraints and optimization objectives.
Rebalancer uses C++ for its core logic and exposes interfaces through Python bindings.
Rebalancer emphasizes 'rebalancing', that is, on the basis of an existing placement, using a solver to find a better allocation scheme, rather than allocating from scratch.
The first step in using Rebalancer is to build a data center model, define the data center's hierarchical structure, and place tasks onto different racks.
Rebalancer integrates HiGHS as an optional backend, letting developers trade off between precision and efficiency as needed.
For developers who want to try Rebalancer, it is recommended to start with the official examples, first using the Python bindings to build a small-scale scenario and understand the complete workflow of modeling, constraints, and solving.
All Facts (7)
Meta's open-source Rebalancer is a solver for data center task placement and optimization, providing a C++ core and Python bindings.
50%UnverifiedRebalancer's core idea is to assign tasks to racks and data centers while satisfying various constraints and optimization objectives.
50%UnverifiedRebalancer uses C++ for its core logic and exposes interfaces through Python bindings.
50%UnverifiedRebalancer emphasizes 'rebalancing', that is, on the basis of an existing placement, using a solver to find a better allocation scheme, rather than allocating from scratch.
50%UnverifiedThe first step in using Rebalancer is to build a data center model, define the data center's hierarchical structure, and place tasks onto different racks.
50%UnverifiedRebalancer integrates HiGHS as an optional backend, letting developers trade off between precision and efficiency as needed.
50%UnverifiedFor developers who want to try Rebalancer, it is recommended to start with the official examples, first using the Python bindings to build a small-scale scenario and understand the complete workflow of modeling, constraints, and solving.
50%