Backend and network systems for real-time media.

KONG Xiao

Software Engineer (Backend)

I am a backend software engineer based in Hangzhou. My work focuses on latency-critical networked systems, especially transport, congestion control, loss recovery, and production observability for real-time streaming.

Before joining Alibaba Cloud, I worked on Tencent START Cloud Gaming, where I shipped transport and telemetry-driven optimization systems and co-authored papers at NSDI. I received my M.S. from Tsinghua University and B.Eng. from Nankai University.

WAN transportReal-time media streamingBackend systemsCloud gaming infrastructure

View full bio Browse papers

Recent highlights

Joined Alibaba Cloud on 2026-03-16.

Returned from a 75-day round-the-world trip on 2025-11-15.

LADR was accepted to NSDI '26 Spring on 2025-07-13 UTC.

Selected papers

See all

From Source to Solution: Tackling Packet Losses in Large-scale Cloud Gaming Systematically and Precisely

23rd USENIX Symposium on Networked Systems Design and Implementation (NSDI '26 Spring)2026

From Source to Solution: Tackling Packet Losses in Large-scale Cloud Gaming Systematically and Precisely

Cloud gaming requires video frames to arrive before strict delay deadlines, yet packet losses can magnify frame delay and nullify the benefits of existing recovery schemes. This work studies packet losses on Tencent START Cloud Gaming, identifies avoidable loss sources and inefficient loss detection, and presents LADR, a combined loss avoidance, detection, and recovery framework. LADR has been deployed on Tencent START for one year and delivers 99.87% of video frames within 100 ms.

cloud gamingpacket lossreal-time streaming

Hairpin: Rethinking Packet Loss Recovery in Edge-based Interactive Video Streaming

21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24)2024

Hairpin: Rethinking Packet Loss Recovery in Edge-based Interactive Video Streaming

Hairpin is a packet loss recovery mechanism for edge-based interactive streaming. It jointly optimizes data packets, retransmissions, and redundant packets across multiple transmission rounds, reducing both bandwidth cost and deadline miss rate for production interactive video workloads.

packet lossinteractive videoedge

Pudica: Toward Near-Zero Queuing Delay in Congestion Control for Cloud Gaming

21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24)2024

Pudica: Toward Near-Zero Queuing Delay in Congestion Control for Cloud Gaming

Pudica is an end-to-end congestion control algorithm for cloud gaming that targets near-zero queuing delay, high link utilization, and cross-flow fairness. In real-network evaluations, it reduces average and tail frame delay and has been deployed in a large-scale cloud gaming platform.

congestion controlcloud gamingreal-time streaming