Rohan Gandhi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Rohan Gandhi has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Rahul Bothra
UIUC, Urbana, USA
3 shared publications - Venkata N. Padmanabhan
Microsoft Research, Bangalore, India
3 shared publications - Debopam Bhattacherjee
Microsoft Research, Bangalore, India
2 shared publications - Ken Sueda
Microsoft, Redmond, USA
2 shared publications - Ranjita Bhagwan
Microsoft Research India amd Google
2 shared publications - Rui Liang
Microsoft, Redmond, USA
2 shared publications
Latest Publications
- Open Access
- 17 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 134-141
FaultSense: Fault Localization in Large-Scale Mixture-of-Experts Model Serving Infrastructure
While Mixture-of-Experts (MoE) models scale LLM serving across hundreds of GPUs, their reliance on all-to-all communication makes them susceptible to gray failures (e.g., GPU stragglers) which inflate serving latency without explicit errors, complicating fault localization. We present FaultSense, an application-layer approach that localizes faulty GPUs and communication paths in MoE model serving replicas without host instrumentation. FaultSense introduces a lightweight MoE probe model prototype alongside a two-phase hierarchical fault localization algorithm. Modeling the cluster as a GPU communication graph, it performs: (1) a global coverage that tests all GPUs and communication paths with a small set of probes to identify suspicious components, and then (2) a drill-down into suspicious components to localize faulty GPUs or communication paths. Our design reduces the number of diagnostic tests compared to individually testing all components (up to 20×) while maintaining low memory overheads.
- 17 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 134-141
Open AccessFaultSense: Fault Localization in Large-Scale Mixture-of-Experts Model Serving Infrastructure
While Mixture-of-Experts (MoE) models scale LLM serving across hundreds of GPUs, their reliance on all-to-all communication makes them susceptible to gray failures (e.g., GPU stragglers) which inflate serving latency without explicit errors, complicating fault localization. We present FaultSense, an application-layer approach that localizes faulty GPUs and communication paths in MoE model serving replicas without host instrumentation. FaultSense introduces a lightweight MoE probe model prototype alongside a two-phase hierarchical fault localization algorithm. Modeling the cluster as a GPU communication graph, it performs: (1) a global coverage that tests all GPUs and communication paths with a small set of probes to identify suspicious components, and then (2) a drill-down into suspicious components to localize faulty GPUs or communication paths. Our design reduces the number of diagnostic tests compared to individually testing all components (up to 20×) while maintaining low memory overheads.
- Open Access
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 183-195
P n M: Efficient Intra-Datacenter Calls Packing for Large Conferencing Services
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 183-195
Open AccessP n M: Efficient Intra-Datacenter Calls Packing for Large Conferencing Services
- Open Access
- 5 March 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Networking
- Vol. 3 (CoNEXT1) , 1-20
KnapsackLB: Enabling Performance-Aware Layer-4 Load Balancing
The layer-4 load balancer (LB) is one of the key building blocks of online services. In this paper, we empower such LBs to adapt to different and dynamic performance of backend instances (DIPs). Our system, KnapsackLB, is generic (can work with variety of LBs), does not require agents on DIPs, LBs or clients, and scales to large numbers of DIPs. KnapsackLB uses judicious active probes to learn a mapping from LB weights to the response latency of each DIP, and then applies Integer Linear Programming (ILP) to calculate LB weights that optimize latency, using an iterative method to scale the computation to large numbers of DIPs. Using testbed experiments and simulations, we show that KnapsackLB load balances traffic according to DIP performance and cuts average latency by up to 45% compared to existing designs.
- 5 March 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Networking
- Vol. 3 (CoNEXT1) , 1-20
Open AccessKnapsackLB: Enabling Performance-Aware Layer-4 Load Balancing
The layer-4 load balancer (LB) is one of the key building blocks of online services. In this paper, we empower such LBs to adapt to different and dynamic performance of backend instances (DIPs). Our system, KnapsackLB, is generic (can work with variety of LBs), does not require agents on DIPs, LBs or clients, and scales to large numbers of DIPs. KnapsackLB uses judicious active probes to learn a mapping from LB weights to the response latency of each DIP, and then applies Integer Linear Programming (ILP) to calculate LB weights that optimize latency, using an iterative method to scale the computation to large numbers of DIPs. Using testbed experiments and simulations, we show that KnapsackLB load balances traffic according to DIP performance and cuts average latency by up to 45% compared to existing designs.
- Open Access
- 25 November 2024
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Networking
- Vol. 2 (CoNEXT4) , 1-22
Saving Private WAN: Using Internet Paths to Offload WAN Traffic in Conferencing Services
Large-scale video conferencing services incur significant network cost while serving surging global demands. Our work systematically explores the opportunity to offload a fraction of this traffic to the Internet, a cheaper routing option offered already by cloud providers, from WAN without drop in application performance. First, with a large-scale latency measurement study with 3.5 million data points per day spanning 241K source cities and 21 data centers across the globe, we demonstrate that Internet paths perform comparable to or better than the private WAN for parts of the world (e.g., Europe and North America). Next, we present Titan, a live (12+ months) production system that carefully moves a fraction of the conferencing traffic to the Internet using the above observation. Finally, we propose Titan-Next - a research prototype that jointly assigns the conferencing server and routing option (Internet or WAN) for individual calls. With 5 weeks of production data, we show Titan-Next reduces the sum of peak bandwidth on WAN links that defines the operational network cost by up to 61% compared to state-of-the-art baselines.
- 25 November 2024
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Networking
- Vol. 2 (CoNEXT4) , 1-22
Open AccessSaving Private WAN: Using Internet Paths to Offload WAN Traffic in Conferencing Services
Large-scale video conferencing services incur significant network cost while serving surging global demands. Our work systematically explores the opportunity to offload a fraction of this traffic to the Internet, a cheaper routing option offered already by cloud providers, from WAN without drop in application performance. First, with a large-scale latency measurement study with 3.5 million data points per day spanning 241K source cities and 21 data centers across the globe, we demonstrate that Internet paths perform comparable to or better than the private WAN for parts of the world (e.g., Europe and North America). Next, we present Titan, a live (12+ months) production system that carefully moves a fraction of the conferencing traffic to the Internet using the above observation. Finally, we propose Titan-Next - a research prototype that jointly assigns the conferencing server and routing option (Internet or WAN) for individual calls. With 5 weeks of production data, we show Titan-Next reduces the sum of peak bandwidth on WAN links that defines the operational network cost by up to 61% compared to state-of-the-art baselines.
- Open Access
- 28 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 109-116
Don't Forget the User
Network measurement has long focused on the bits and bytes --- low-level network metrics such as latency and throughput, which have the advantage of being objective and directly characterizing the performance of the network. We argue that users provide a rich and largely untapped source of implicit as well as explicit signals that could complement and expand the coverage of traditional methods. Implicit feedback leverages user actions to indirectly infer the network performance and the resulting quality of user experience. Explicit feedback leverages user input, typically provided offline, to expand the reach of network measurement, especially for newer ones. We analyze two scenarios: capturing implicit feedback through user actions from a large-scale conferencing service -- MS Teams and gathering explicit feedback via social media posts pertaining to the SpaceX Starlink Low Earth Orbit (LEO) satellite network undergoing deployment. We believe our techniques complement the traditional measurement methods and open up a broad set of research directions, ranging from rethinking measurement tools to designing user-centric networked systems and applications.
- 28 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 109-116
Open AccessDon't Forget the User
Network measurement has long focused on the bits and bytes --- low-level network metrics such as latency and throughput, which have the advantage of being objective and directly characterizing the performance of the network. We argue that users provide a rich and largely untapped source of implicit as well as explicit signals that could complement and expand the coverage of traditional methods. Implicit feedback leverages user actions to indirectly infer the network performance and the resulting quality of user experience. Explicit feedback leverages user input, typically provided offline, to expand the reach of network measurement, especially for newer ones. We analyze two scenarios: capturing implicit feedback through user actions from a large-scale conferencing service -- MS Teams and gathering explicit feedback via social media posts pertaining to the SpaceX Starlink Low Earth Orbit (LEO) satellite network undergoing deployment. We believe our techniques complement the traditional measurement methods and open up a broad set of research directions, ranging from rethinking measurement tools to designing user-centric networked systems and applications.
- Open Access
- 10 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 1000-1011
Switchboard: Efficient Resource Management for Conferencing Services
- 10 September 2023
- Published by Association for Computing Machinery (ACM)
- p. 1000-1011
Open AccessSwitchboard: Efficient Resource Management for Conferencing Services
- Open Access
- 1 June 2011
- Published by Association for Computing Machinery (ACM)
- p. 27-32
The impact of inter-layer network coding on the relative performance of MRC/MDC WiFi media delivery
A primary challenge in multicasting video in a wireless LAN is to deal with the client diversity -- clients may have different channel characteristics and hence receive different numbers of transmissions from the AP. A promising approach to overcome this problem is to combine scalable video coding techniques such as MRC or MDC, which divide a video stream into multiple substreams, with inter-layer network coding. The fundamental challenge in such an approach is to determine the strategy of coding the packets across different layers that maximizes the number of decoded layers at all clients. In [7], the authors showed that inter-layer NC indeed helps the delivery of MRC coded media over the WiFi, and proposed how to efficiently search for the optimal coding strategies online. In this paper, we study (1) how NC can help with WiFi delivery of MDC media, and (2) in particular, due to the different decoding requirements of MDC from MRC, whether WiFi delivery of MDC media can benefit more from NC compared to that of MRC media. Our simulation results are somewhat surprising. Even though MDC is generally shown to outperform MRC in lossy channels, most of the benefit of MDC over MRC is lost after applying NC to both schemes.
- 1 June 2011
- Published by Association for Computing Machinery (ACM)
- p. 27-32
Open AccessThe impact of inter-layer network coding on the relative performance of MRC/MDC WiFi media delivery
A primary challenge in multicasting video in a wireless LAN is to deal with the client diversity -- clients may have different channel characteristics and hence receive different numbers of transmissions from the AP. A promising approach to overcome this problem is to combine scalable video coding techniques such as MRC or MDC, which divide a video stream into multiple substreams, with inter-layer network coding. The fundamental challenge in such an approach is to determine the strategy of coding the packets across different layers that maximizes the number of decoded layers at all clients. In [7], the authors showed that inter-layer NC indeed helps the delivery of MRC coded media over the WiFi, and proposed how to efficiently search for the optimal coding strategies online. In this paper, we study (1) how NC can help with WiFi delivery of MDC media, and (2) in particular, due to the different decoding requirements of MDC from MRC, whether WiFi delivery of MDC media can benefit more from NC compared to that of MRC media. Our simulation results are somewhat surprising. Even though MDC is generally shown to outperform MRC in lossy channels, most of the benefit of MDC over MRC is lost after applying NC to both schemes.