Xin Zhe Khooi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xin Zhe Khooi has a total of 51 co-authors. Here are the top co-authors by number of shared publications:
- Mun Choon Chan
National University of Singapore, Singapore, Singapore
12 shared publications - Cha Hwan Song
National University of Singapore, Singapore, Singapore
5 shared publications - Satis Kumar Permal
National University of Singapore, Singapore, Singapore
4 shared publications - Anuj Kalia
Microsoft, Redmond, USA
2 shared publications - Archit Bhatnagar
University of Michigan, Ann Arbor, Michigan, USA
2 shared publications - Biqing Qiu
GovTech Singapore, Singapore, Singapore
2 shared publications
Latest Publications
- Open Access
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 637-652
T hor : Fine-grained Per-UE Baseband Routing in vRAN
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 637-652
Open AccessT hor : Fine-grained Per-UE Baseband Routing in vRAN
- Open Access
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1849-1851
Demo: W eaver - A System for AI-RAN Compute Sharing with Foundation Model Training
We present Weaver, a system that enables opportunistic foundation model (FM) training alongside realtime-sensitive RAN workloads on GPU-accelerated AI-RAN compute infrastructure. Weaver safely harvests spare compute and enables GPU compute sharing between RAN and FM training workloads at a fine-granularity. Weaver achieves this through a RAN-first design, by introducing (1) a RAN-centric spare compute controller integrated with the MAC scheduler to prioritize RAN performance while stabilizing GPU spare compute, and (2) an elastic, RAN-slot-bounded training policy that decomposes FM training into fine-grained work units and executes only the work that fits within each slot's available spare compute. In this demo, we showcase how Weaver allows RAN and training workloads to safely co-exist with each other within a single cell site without violating RAN deadlines, while allowing training to effectively harvest the available spare compute at a per-slot granularity, substantially improving the training throughput.
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1849-1851
Open AccessDemo: W eaver - A System for AI-RAN Compute Sharing with Foundation Model Training
We present Weaver, a system that enables opportunistic foundation model (FM) training alongside realtime-sensitive RAN workloads on GPU-accelerated AI-RAN compute infrastructure. Weaver safely harvests spare compute and enables GPU compute sharing between RAN and FM training workloads at a fine-granularity. Weaver achieves this through a RAN-first design, by introducing (1) a RAN-centric spare compute controller integrated with the MAC scheduler to prioritize RAN performance while stabilizing GPU spare compute, and (2) an elastic, RAN-slot-bounded training policy that decomposes FM training into fine-grained work units and executes only the work that fits within each slot's available spare compute. In this demo, we showcase how Weaver allows RAN and training workloads to safely co-exist with each other within a single cell site without violating RAN deadlines, while allowing training to effectively harvest the available spare compute at a per-slot granularity, substantially improving the training throughput.
- Open Access
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1816-1818
Demo: Per-UE Baseband Routing in vRAN
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1816-1818
Open AccessDemo: Per-UE Baseband Routing in vRAN
- Open Access
- 26 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 263-278
Handling Network Faults in Distributed AI Training: Failover is Now an Option
- 26 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 263-278
Open AccessHandling Network Faults in Distributed AI Training: Failover is Now an Option
- Open Access
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 123-138
How to Update Your 5G vRAN
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 123-138
Open AccessHow to Update Your 5G vRAN
- Open Access
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1216-1218
Demo: Towards Seamless 5G vRAN Software Updates
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1216-1218
Open AccessDemo: Towards Seamless 5G vRAN Software Updates
- Open Access
- 16 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 385-397
Revisiting CPU Performance Scaling for Energy-Efficient Packet Processing Applications
- 16 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 385-397
Open AccessRevisiting CPU Performance Scaling for Energy-Efficient Packet Processing Applications
- Open Access
- 9 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 14-20
Towards Continuous Latency Monitoring through Open RAN Interfaces
We propose DBench, a framework for automated data collection and analysis to infer the processing latency of individual RAN components. DBench interposes on open RAN interfaces, periodically collecting user data traffic and correlating them to provide a detailed breakdown of user-perceived network latency. Developing DBench involves addressing challenges such as mapping low-level raw I/Q samples to individual data packets and handling of high volume of RAN traffic. We overcome these challenges by (1) exploiting the timing characteristics of the 5G New Radio (NR) frame structure at the fronthaul interface, and (2) implementing a pipeline data collection and parallel data processing flow. We demonstrate DBench's capabilities through several use cases on an academic open RAN testbed and discuss ongoing efforts to expand its functionalities and address current limitations.
- 9 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 14-20
Open AccessTowards Continuous Latency Monitoring through Open RAN Interfaces
We propose DBench, a framework for automated data collection and analysis to infer the processing latency of individual RAN components. DBench interposes on open RAN interfaces, periodically collecting user data traffic and correlating them to provide a detailed breakdown of user-perceived network latency. Developing DBench involves addressing challenges such as mapping low-level raw I/Q samples to individual data packets and handling of high volume of RAN traffic. We overcome these challenges by (1) exploiting the timing characteristics of the 5G New Radio (NR) frame structure at the fronthaul interface, and (2) implementing a pipeline data collection and parallel data processing flow. We demonstrate DBench's capabilities through several use cases on an academic open RAN testbed and discuss ongoing efforts to expand its functionalities and address current limitations.
- Open Access
- 6 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 41-48
P4EAD: Securing the In-band Control Channels on Commodity Programmable Switches
Conventionally, the control channel on network switches has always been out-of-band. With the emergence of high-performance systems built upon programmable switches, the out-of-band control channel has become the bottleneck. Thus, there is an emerging trend of implementing the control channel in the data path (i.e., in-band) on programmable switches to achieve high throughput and low-latency control actions. However, the use of in-band control channels comes with the risk of security vulnerabilities that have not been explored in prior literature. In this paper, we present P4EAD, a cryptographic primitive to secure the in-band control channels on programmable switches entirely in the data plane. This ensures the integrity, authenticity, and confidentiality of in-band control messages. We conduct micro-benchmarks on P4EAD and demonstrate its integration with an existing high-performance in-band control framework, showcasing minimal performance impact when securing the control channel.
- 6 December 2023
- Published by Association for Computing Machinery (ACM)
- p. 41-48
Open AccessP4EAD: Securing the In-band Control Channels on Commodity Programmable Switches
Conventionally, the control channel on network switches has always been out-of-band. With the emergence of high-performance systems built upon programmable switches, the out-of-band control channel has become the bottleneck. Thus, there is an emerging trend of implementing the control channel in the data path (i.e., in-band) on programmable switches to achieve high throughput and low-latency control actions. However, the use of in-band control channels comes with the risk of security vulnerabilities that have not been explored in prior literature. In this paper, we present P4EAD, a cryptographic primitive to secure the in-band control channels on programmable switches entirely in the data plane. This ensures the integrity, authenticity, and confidentiality of in-band control messages. We conduct micro-benchmarks on P4EAD and demonstrate its integration with an existing high-performance in-band control framework, showcasing minimal performance impact when securing the control channel.
- Open Access
- 10 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 1150-1152
Poster: Towards Accelerating the 5G Centralized Unit with Programmable Switches
- 10 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 1150-1152
Open AccessPoster: Towards Accelerating the 5G Centralized Unit with Programmable Switches