Network testers are essential for assessing network performance. They are used to generate and capture test packets to evaluate the network's correctness and efficiency. However, evolving demands, such as generating realistic, high-performance workloads, pose challenges for existing solutions. In this demonstration, we present P4 replay (P4R), a network tester based on a programmable switch ASIC capable of reproducing real packet traces and establishing high-performance stateful TCP connections. P4R can test external clients and servers with high throughput and accuracy as well as self-testing P4 applications running in parallel on the same device.
Recent advancements in video streaming technologies have brought about a new era of user experiences, including gaming and online events. However, ensuring a good user experience remains a significant challenge. The pursuit of self-driving networks has increased the demand for intelligent, edge-based networking solutions. Delivering high-quality video streaming experiences faces hurdles such as varying network conditions, device diversity, and content complexity, along with the inter-twinned adaptivity of over-the-top services and modern network management systems. In this work, we elaborate on video streaming quality of experience (QoE) improvements based on recent advances in programmable networks. We explore the concept of in-network QoE (INQoE) by discussing benefits, challenges, and opportunities. Furthermore, we present a pure data plane strategy to estimate and improve the QoE of 360-degree virtual reality video streaming. Our results show an INQoE management strategy can effectively keep the QoE close to optimal (4.8 on a scale of 5) under a congestion scenario. Finally, we conclude with a roadmap of research challenges and opportunities around INQoE.
Ensuring bounded end-to-end latency and high reliability is essential for time-critical services, motivating the development of TSN and DetNet protocols to guarantee predictable and reliable network performance. Testing the performance of IEEE and IETF standards can be challenging due to the cost of commercial devices supporting their implementation. In this work, we present TFTG, a traffic generator based on Tofino capable of creating TSN and DetNet traffic with nanoscale precision and high throughput. The proposed solution can be used in different scenarios to test standards for synchronization, reliability, and scheduling, with high flexibility in its applications. With these functionalities, TFTG serves as an open-source solution that users can employ to test target device capabilities, abstracting different topologies and scenarios for multiple protocols and testbed configurations.
Intra-server resource orchestration is becoming increasingly challenging. Modern server architectures are now composed of multiple computing units, such as CPUs and DPUs. VNFs should be appropriately provisioned between these computing units, avoiding performance degradation and optimizing server resources. In this work, we introduce In-Network Resource Allocation (InReal), a system to orchestrate containerized VNFs. InReal can provision and manage VNF resources, optimizing CPU usage and enhancing power efficiency while meeting VNF performance requirements.