qUpdate: QoS-aware Consistent Updates in SDN

A significant volume of work exists to handle consistent updates in Software Defined Networks (SDN). Often the guarantee of consistency sacrifices performance leading to Quality of Service (QoS) violations in real settings. The challenge of preserving QoS at all times stems from the unknown duration of each update. Given an ordering of updates, the specific intermediary states of the network during a set of updates are unknown until run-time and can violate QoS. To address this problem, we propose qUpdate, the first QoS-aware consistent update framework, to enable fast network updates in SDN while ensuring QoS requirements are met at all times. Specifically, qUpdate introduces a novel transformation to model the behaviours of a network as a set of Boolean constraints. These are combined with QoS and resource constraints as inputs for an SMT solver to compute all orderings where all possible intermediary states are QoS compliant. To best harness this set, we introduce a geometric model to dynamically choose the optimal ordering as each update completes and leads to a new intermediary state. We implement the qUpdate prototype and conduct extensive benchmarking using a real controller. The evaluation results show that qUpdate can effectively perform QoS-compliant network updates, and speed up the updates by up to 2.5x compared to prior work.