TWRR

Time-Weighted Round Robin (TWRR) is a resource scheduling algorithm used to allocate processing time or bandwidth among multiple competing queues or tasks. It is a variant of Weighted Round Robin that incorporates time-aware adjustments to scheduling quotas, aiming to provide fair shares even when tasks vary in size or service requirements.

In typical implementations, each queue i is assigned a weight w_i that represents its priority or desired share. The scheduler cycles through nonempty queues and assigns service time to each in proportion to its weight, with adjustments made based on observed service times so that longer tasks do not disproportionately penalize others. A virtual time reference is often used to track progress and to decide the next queue to serve, ensuring the cumulative service delivered matches the weighted shares over time.

Applications for TWRR include quality of service in network devices such as routers and switches, where different traffic classes receive different fractions of bandwidth, and operating system or virtualization schedulers that must fairly multiplex CPU or I/O time among processes or virtual machines.

Advantages of TWRR include improved fairness for mixed workloads and easier integration with existing weighted round-robin frameworks. Limitations involve the need to select appropriate weights, potential performance variations under bursty traffic, and the fact that implementations can differ across systems, leading to behavior that is not fully standardized.

TWRR is not universally standardized; it exists as a family of implementations described in vendor-specific QoS or scheduling documentation. It is related to WRR (weighted round robin), DRR (deficit round robin), and fair queuing approaches.