This paper is currently under review in MobiCom 2025

This paper is currently under review in IEEE/ACM Transactions on Networking

5G networks are considered potential enablers for many emerging edge applications, such as those related to autonomous vehicles, virtual and augmented reality, and online gaming. However, recent works have shown the cellular control plane is a potential bottleneck in enabling such applications — control plane operations are slow, frequent, and can directly impact the delay experienced by end-user applications. Moreover, failures in the cellular control plane can significantly degrade application performance. In this paper, we consider the problem of enabling latency-sensitive and safety-critical edge applications on 5G networks. We identify fundamental control plane design challenges and posit enabling these applications requires re-thinking the cellular control plane. We propose a new edge-based cellular control plane, CellClone, which provides fast and fault-tolerant control plane processing. CellClone employs multiple active control plane clones at the network edge to mask control plane faults and speed up control processing. Central to its design is a custom quorum-based consistency protocol that provides state consistency with low latency. Testbed evaluations using real cellular traces show a median improvement of more than 3.8× in speeding up control plane operations with outright node failures and stragglers. These improvements translate into better application performance; with CellClone, autonomous cars and VR applications reduce missed application deadlines by more than 90%.

5G and next-generation cellular networks aim to support tactile internet to enable immersive and real-time applications by providing ultra-low latency and extremely high reliability. This imposes new requirements on the design of cellular core networks. A key component of the cellular core is the control plane. Time to complete cellular control plane operations (e.g., mobility handoff, service establishment) directly impacts the delay experienced by end-user applications. In this paper, we design Neutrino, a cellular control plane that provides users an abstraction of reliable access to cellular services while ensuring lower latency. Our testbed evaluations based on real cellular control traffic traces show that Neutrino provides an improvement in control procedure completion times by up to 3.1× without failures, and up to 5.6× under control plane failures, over existing 5G. We also show how these improvements translate into improving end-user application performance: for AR/VR applications and self-driving cars, Neutrino improves performance by up to 2.5× and 2.8×, respectively.

5G networks aim to provide ultra-low latency and higher reliability to support emerging and near real-time applications such as augmented and virtual reality, remote surgery, self-driving cars, and multi-player online gaming. This imposes new requirements on the design of cellular core networks. A key component of the cellular core is the control plane. Time to complete control plane operations (e.g. mobility handoff, service establishment) directly impacts the delay experienced by end-user applications. In this paper, we design Neutrino, a cellular control plane that provides users an abstraction of reliable access to cellular services while ensuring lower latency. Our testbed evaluations based on real cellular control traffic traces show that Neutrino provides an improvement in control procedure completion times by up to 3.1x without failures, and up to 5.6x under control plane failures, over existing cellular core proposals. We also show how these improvements translate into improving end-user application performance: for AR/VR applications and self-driving cars, Neutrino performs up to 2.5x and up to 2.8x better, respectively, as compared to existing EPC.

Timely completion of control plane operations is critical to providing fast data access in cellular networks. In this work, we design a new edge-based cellular control plane, Fast EPC, which reduces control plane latency through fast serialization of control messages and rapid failure recovery.