Subho Sankar Banerjee
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Subho Sankar Banerjee has a total of 51 co-authors. Here are the top co-authors by number of shared publications:
Ravishankar K. Iyer
University of Illinois at Urbana-Champaign
13 shared publicationsZbigniew T. Kalbarczyk
University of Illinois at Urbana-Champaign
10 shared publicationsSaurabh Jha
6 shared publicationsJames Cyriac
University of Illinois at Urbana-Champaign
4 shared publicationsArjun P. Athreya
University of Illinois at Urbana-Champaign, Urbana IL
3 shared publicationsCostin Grigoras
2 shared publications
Latest Publications
- Open Access
- 24 August 2020
- Published by Association for Computing Machinery (ACM)
- p. 67-74
Machine learning for load balancing in the Linux kernel
The OS load balancing algorithm governs the performance gains provided by a multiprocessor computer system. The Linux's Completely Fair Scheduler (CFS) scheduler tracks process loads by average CPU utilization to balance workload between processor cores. That approach maximizes the utilization of processing time but overlooks the contention for lower-level hardware resources. In servers running compute-intensive workloads, an imbalanced need for limited computing resources hinders execution performance. This paper solves the above problem using a machine learning (ML)-based resource-aware load balancer. We describe (1) low-overhead methods for collecting training data; (2) an ML model based on a multi-layer perceptron model that imitates the CFS load balancer based on the collected training data; and (3) an in-kernel implementation of inference on the model. Our experiments demonstrate that the proposed model has an accuracy of 99% in making migration decisions and while only increasing the latency by 1.9 μs.
- 24 August 2020
- Published by Association for Computing Machinery (ACM)
- p. 67-74
Open AccessMachine learning for load balancing in the Linux kernel
The OS load balancing algorithm governs the performance gains provided by a multiprocessor computer system. The Linux's Completely Fair Scheduler (CFS) scheduler tracks process loads by average CPU utilization to balance workload between processor cores. That approach maximizes the utilization of processing time but overlooks the contention for lower-level hardware resources. In servers running compute-intensive workloads, an imbalanced need for limited computing resources hinders execution performance. This paper solves the above problem using a machine learning (ML)-based resource-aware load balancer. We describe (1) low-overhead methods for collecting training data; (2) an ML model based on a multi-layer perceptron model that imitates the CFS load balancer based on the collected training data; and (3) an in-kernel implementation of inference on the model. Our experiments demonstrate that the proposed model has an accuracy of 99% in making migration decisions and while only increasing the latency by 1.9 μs.
- Open Access
Inductive-bias-driven Reinforcement Learning For Efficient Schedules in Heterogeneous Clusters
The problem of scheduling of workloads onto heterogeneous processors (e.g., CPUs, GPUs, FPGAs) is of fundamental importance in modern data centers. Current system schedulers rely on application/system-specific heuristics that have to be built on a case-by-case basis. Recent work has demonstrated ML techniques for automating the heuristic search by using black-box approaches which require significant training data and time, which make them challenging to use in practice. This paper presents Symphony, a scheduling framework that addresses the challenge in two ways: (i) a domain-driven Bayesian reinforcement learning (RL) model for scheduling, which inherently models the resource dependencies identified from the system architecture; and (ii) a sampling-based technique to compute the gradients of a Bayesian model without performing full probabilistic inference. Together, these techniques reduce both the amount of training data and the time required to produce scheduling policies that significantly outperform black-box approaches by up to 2.2x.
Open AccessInductive-bias-driven Reinforcement Learning For Efficient Schedules in Heterogeneous Clusters
The problem of scheduling of workloads onto heterogeneous processors (e.g., CPUs, GPUs, FPGAs) is of fundamental importance in modern data centers. Current system schedulers rely on application/system-specific heuristics that have to be built on a case-by-case basis. Recent work has demonstrated ML techniques for automating the heuristic search by using black-box approaches which require significant training data and time, which make them challenging to use in practice. This paper presents Symphony, a scheduling framework that addresses the challenge in two ways: (i) a domain-driven Bayesian reinforcement learning (RL) model for scheduling, which inherently models the resource dependencies identified from the system architecture; and (ii) a sampling-based technique to compute the gradients of a Bayesian model without performing full probabilistic inference. Together, these techniques reduce both the amount of training data and the time required to produce scheduling policies that significantly outperform black-box approaches by up to 2.2x.
- 1 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 113-124
ML-Driven Malware that Targets AV Safety
Ensuring the safety of autonomous vehicles (AVs) is critical for their mass deployment and public adoption. However, security attacks that violate safety constraints and cause accidents are a significant deterrent to achieving public trust in AVs, and that hinders a vendor's ability to deploy AVs. Creating a security hazard that results in a severe safety compromise (for example, an accident) is compelling from an attacker's perspective. In this paper, we introduce an attack model, a method to deploy the attack in the form of smart malware, and an experimental evaluation of its impact on production-grade autonomous driving software. We find that determining the time interval during which to launch the attack is{ critically} important for causing safety hazards (such as collisions) with a high degree of success. For example, the smart malware caused 33X more forced emergency braking than random attacks did, and accidents in 52.6% of the driving simulations.
- 1 June 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 113-124
ML-Driven Malware that Targets AV Safety
Ensuring the safety of autonomous vehicles (AVs) is critical for their mass deployment and public adoption. However, security attacks that violate safety constraints and cause accidents are a significant deterrent to achieving public trust in AVs, and that hinders a vendor's ability to deploy AVs. Creating a security hazard that results in a severe safety compromise (for example, an accident) is compelling from an attacker's perspective. In this paper, we introduce an attack model, a method to deploy the attack in the form of smart malware, and an experimental evaluation of its impact on production-grade autonomous driving software. We find that determining the time interval during which to launch the attack is{ critically} important for causing safety hazards (such as collisions) with a high degree of success. For example, the smart malware caused 33X more forced emergency braking than random attacks did, and accidents in 52.6% of the driving simulations.
- 1 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 112-124
ML-Based Fault Injection for Autonomous Vehicles: A Case for Bayesian Fault Injection
The safety and resilience of fully autonomous vehicles (AVs) are of significant concern, as exemplified by several headline-making accidents. While AV development today involves verification, validation, and testing, end-to-end assessment of AV systems under accidental faults in realistic driving scenarios has been largely unexplored. This paper presents DriveFI, a machine learning-based fault injection engine, which can mine situations and faults that maximally impact AV safety, as demonstrated on two industry-grade AV technology stacks (from NVIDIA and Baidu). For example, DriveFI found 561 safety-critical faults in less than 4 hours. In comparison, random injection experiments executed over several weeks could not find any safety-critical faults.
- 1 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 112-124
ML-Based Fault Injection for Autonomous Vehicles: A Case for Bayesian Fault Injection
The safety and resilience of fully autonomous vehicles (AVs) are of significant concern, as exemplified by several headline-making accidents. While AV development today involves verification, validation, and testing, end-to-end assessment of AV systems under accidental faults in realistic driving scenarios has been largely unexplored. This paper presents DriveFI, a machine learning-based fault injection engine, which can mine situations and faults that maximally impact AV safety, as demonstrated on two industry-grade AV technology stacks (from NVIDIA and Baidu). For example, DriveFI found 561 safety-critical faults in less than 4 hours. In comparison, random injection experiments executed over several weeks could not find any safety-critical faults.
- 1 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 25-26
Towards a Bayesian Approach for Assessing Fault Tolerance of Deep Neural Networks
This paper presents Bayesian Deep Learning based Fault Injection (BDLFI), a novel methodology for fault injection in neural networks (NNs) and more generally differentiable programs. BDLFI uses (1) Bayesian Deep Learning to model the propagation of faults, and (2) Markov Chain Monte Carlo inference to quantify the effect of faults on the outputs of a NN. We demonstrate BDLFI on two representative networks and present our results that challenge pre-existing results in the field.
- 1 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 25-26
Towards a Bayesian Approach for Assessing Fault Tolerance of Deep Neural Networks
This paper presents Bayesian Deep Learning based Fault Injection (BDLFI), a novel methodology for fault injection in neural networks (NNs) and more generally differentiable programs. BDLFI uses (1) Bayesian Deep Learning to model the propagation of faults, and (2) Markov Chain Monte Carlo inference to quantify the effect of faults on the outputs of a NN. We demonstrate BDLFI on two representative networks and present our results that challenge pre-existing results in the field.
- 12 October 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 68 (3) , 331-346
ASAP: Accelerated Short-Read Alignment on Programmable Hardware
The proliferation of high-throughput sequencing machines ensures rapid generation of up to billions of short nucleotide fragments in a short period of time. This massive amount of sequence data can quickly overwhelm today's storage and compute infrastructure. This paper explores the use of hardware acceleration to significantly improve the runtime of short-read alignment, a crucial step in preprocessing sequenced genomes. We focus on the Levenshtein distance (edit-distance) computation kernel and propose the ASAP accelerator, which utilizes the intrinsic delay of circuits for edit-distance computation elements as a proxy for computation. Our design is implemented on an Xilinx Virtex 7 FPGA in an IBM POWER8 system that uses the CAPI interface for cache coherence across the CPU and FPGA. Our design is 200x faster than an equivalent Smith-Waterman-C implementation of the kernel running on the host processor, 40-60x faster than an equivalent Landau-Vishkin-C++ implementation of the kernel running on the IBM Power8 host processor, and 2x faster for an end-to-end alignment tool for 120--150 base-pair short-read sequences. Further the design represents a 3760x improvement over the CPU in performance/Watt terms.
- 12 October 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 68 (3) , 331-346
ASAP: Accelerated Short-Read Alignment on Programmable Hardware
The proliferation of high-throughput sequencing machines ensures rapid generation of up to billions of short nucleotide fragments in a short period of time. This massive amount of sequence data can quickly overwhelm today's storage and compute infrastructure. This paper explores the use of hardware acceleration to significantly improve the runtime of short-read alignment, a crucial step in preprocessing sequenced genomes. We focus on the Levenshtein distance (edit-distance) computation kernel and propose the ASAP accelerator, which utilizes the intrinsic delay of circuits for edit-distance computation elements as a proxy for computation. Our design is implemented on an Xilinx Virtex 7 FPGA in an IBM POWER8 system that uses the CAPI interface for cache coherence across the CPU and FPGA. Our design is 200x faster than an equivalent Smith-Waterman-C implementation of the kernel running on the host processor, 40-60x faster than an equivalent Landau-Vishkin-C++ implementation of the kernel running on the IBM Power8 host processor, and 2x faster for an end-to-end alignment tool for 120--150 base-pair short-read sequences. Further the design represents a 3760x improvement over the CPU in performance/Watt terms.
- Open Access
- 11 October 2018
- Published by Association for Computing Machinery (ACM)
- p. 533-534
A ML-based Runtime System for Executing Dataflow Graphs on Heterogeneous Processors
- 11 October 2018
- Published by Association for Computing Machinery (ACM)
- p. 533-534
Open AccessA ML-based Runtime System for Executing Dataflow Graphs on Heterogeneous Processors
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 586-597
Hands Off the Wheel in Autonomous Vehicles?: A Systems Perspective on over a Million Miles of Field Data
Autonomous vehicle (AV) technology is rapidly becoming a reality on U.S. roads, offering the promise of improvements in traffic management, safety, and the comfort and efficiency of vehicular travel. The California Department of Motor Vehicles (DMV) reports that between 2014 and 2017, manufacturers tested 144 AVs, driving a cumulative 1,116,605 autonomous miles, and reported 5,328 disengagements and 42 accidents involving AVs on public roads. This paper investigates the causes, dynamics, and impacts of such AV failures by analyzing disengagement and accident reports obtained from public DMV databases. We draw several conclusions. For example, we find that autonomous vehicles are 15 - 4000Ã- worse than human drivers for accidents per cumulative mile driven; that drivers of AVs need to be as alert as drivers of non-AVs; and that the AVs' machine-learning-based systems for perception and decision-and-control are the primary cause of 64% of all disengagements.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 586-597
Hands Off the Wheel in Autonomous Vehicles?: A Systems Perspective on over a Million Miles of Field Data
Autonomous vehicle (AV) technology is rapidly becoming a reality on U.S. roads, offering the promise of improvements in traffic management, safety, and the comfort and efficiency of vehicular travel. The California Department of Motor Vehicles (DMV) reports that between 2014 and 2017, manufacturers tested 144 AVs, driving a cumulative 1,116,605 autonomous miles, and reported 5,328 disengagements and 42 accidents involving AVs on public roads. This paper investigates the causes, dynamics, and impacts of such AV failures by analyzing disengagement and accident reports obtained from public DMV databases. We draw several conclusions. For example, we find that autonomous vehicles are 15 - 4000Ã- worse than human drivers for accidents per cumulative mile driven; that drivers of AVs need to be as alert as drivers of non-AVs; and that the AVs' machine-learning-based systems for perception and decision-and-control are the primary cause of 64% of all disengagements.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 55-56
AVFI: Fault Injection for Autonomous Vehicles
Autonomous vehicle (AV) technology is rapidly becoming a reality on U.S. roads, offering the promise of improvements in traffic management, safety, and the comfort and efficiency of vehicular travel. With this increasing popularity and ubiquitous deployment, resilience has become a critical requirement for public acceptance and adoption. Recent studies into the resilience of AVs have shown that though the AV systems are improving over time, they have not reached human levels of automation. Prior work in this area has studied the safety and resilience of individual components of the AV system (e.g., testing of neural networks powering the perception function). However, methods for holistic end-to-end resilience assessment of AV systems are still non-existent.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 55-56
AVFI: Fault Injection for Autonomous Vehicles
Autonomous vehicle (AV) technology is rapidly becoming a reality on U.S. roads, offering the promise of improvements in traffic management, safety, and the comfort and efficiency of vehicular travel. With this increasing popularity and ubiquitous deployment, resilience has become a critical requirement for public acceptance and adoption. Recent studies into the resilience of AVs have shown that though the AV systems are improving over time, they have not reached human levels of automation. Prior work in this area has studied the safety and resilience of individual components of the AV system (e.g., testing of neural networks powering the perception function). However, methods for holistic end-to-end resilience assessment of AV systems are still non-existent.