conference paper

SCAR: Scheduling Multi-Model AI Workloads on Heterogeneous Multi-Chiplet Module Accelerators

2024 57th IEEE/ACM International Symposium on Microarchitecture (MICRO)

Publication Date

November 1, 2024

Author(s)

Mohanad Odema, Luke Chen, Hyoukjun Kwon, Mohammad Al Faruque

Abstract

Emerging multi-model workloads with heavy models like recent large language models significantly increased the compute and memory demands on hardware. To address such increasing demands, designing a scalable hardware architecture became a key problem. Among recent solutions, the 2.5D silicon interposer multi-chip module (MCM)-based AI accelerator has been actively explored as a promising scalable solution due to their significant benefits in the low engineering cost and composability. However, previous MCM accelerators are based on homogeneous architectures with fixed dataflow, which encounter major challenges from highly heterogeneous multi-model work-loads due to their limited workload adaptivity. Therefore, in this work, we explore the opportunity in the heterogeneous dataflow MCM AI accelerators. We identify the scheduling of multi-model workload on heterogeneous dataflow MCM AI accelerator is an important and challenging problem due to its significance and scale, which reaches mathbfO(10^56) even for a two-model workload on 6×6 chiplets. We develop a set of heuristics to navigate the huge scheduling space and codify them into a scheduler, SCAR, with advanced techniques such as inter-chiplet pipelining. Our evaluation on ten multi-model workload scenarios for datacenter multitenancy and AR/VR use-cases has shown the efficacy of our approach, achieving on average 27.6% and 29.6% less energy-delay product (EDP) for the respective applications settings compared to homogeneous baselines.

Suggested Citation
Mohanad Odema, Luke Chen, Hyoukjun Kwon and Mohammad Abdullah Al Faruque (2024) “SCAR: Scheduling Multi-Model AI Workloads on Heterogeneous Multi-Chiplet Module Accelerators”, in 2024 57th IEEE/ACM International Symposium on Microarchitecture (MICRO). 2024 57th IEEE/ACM International Symposium on Microarchitecture (MICRO), pp. 565–579. Available at: 10.1109/MICRO61859.2024.00049.

published journal article

Alliances, codesharing, antitrust immunity, and international airfares: Do previous patterns persist?

Journal of Competition Law and Economics

Publication Date

May 1, 2011

Author(s)

Jan Brueckner, Dennis Lee, E.S. Singer
Suggested Citation
J.K. Brueckner, D.N. Lee and E.S. Singer (2011) “Alliances, codesharing, antitrust immunity, and international airfares: Do previous patterns persist?”, Journal of Competition Law and Economics, 7(3), pp. 573–602. Available at: 10.1093/joclec/nhr005.

working paper

Subcenters in the Los Angeles Region

Publication Date

October 30, 1991

Associated Project

Working Paper

No. 39

Areas of Expertise

Abstract

We investigate employment subcenters in the Los Angeles region using 1980 Census journey-to-work data. A simple subcenter definition is used, based solely on gross employment density and total employment. We find a surprising dominance of downtown Los Angeles and three large subcenters with which it forms a nearly contiguous corridor. Two-thirds of the region’s employment, however, is outside any of the 32 centers we identify. Most centers have high population densities in and near them, and their workers’ commutes are just 2.4 miles longer than other workers’ commutes. A cluster analysis of employment by industry reveals several distinct types of centers, and a wide dispersion of sizes and locations within each type.

Suggested Citation
Genevieve Giuliano and Kenneth A. Small (1991) Subcenters in the Los Angeles Region. Working Paper No. 39. Institute of Transportation Studies, UC Irvine: University of California Transportation Center. Available at: https://escholarship.org/uc/item/7xv976dj.

conference paper

Managing residential-level EV charging using network-as-automation platform (NAP) technology

2012 IEEE international electric vehicle conference

Publication Date

March 1, 2012

Author(s)

Mohammad Al Faruque, Livio Dalloro, Siyuan Zhou, Hartmut Ludwig, George Lo
Suggested Citation
Mohammad Abdullah Al Faruque, Livio Dalloro, Siyuan Zhou, Hartmut Ludwig and George Lo (2012) “Managing residential-level EV charging using network-as-automation platform (NAP) technology”, in 2012 IEEE international electric vehicle conference. IEEE, pp. 1–6. Available at: 10.1109/ievc.2012.6183218.

published journal article

A vehicle use forecasting model based on revealed and stated vehicle type choice and utilisation data

Journal of Transport Economics and Policy

Publication Date

January 1, 1997

Author(s)

Abstract

Structural equation models are developed that explain driver allocation and annual vehicle use in terms of characteristics of the household and its vehicle(s). Separate models are estimated for one-vehicle and multi-vehicle households using both observed and stated preference data from a 1993 survey in California. Results quantify how use decreases as vehicles age, and it is predicted that electric and other limited range vehicles will be used less than conventional fuel vehicles. In multi-vehicle households, shifts in use are detected between vehicles of different ages, operating costs and fuel types. /// Es werden Strukturgleichungsmodelle entwickelt, die den jährlichen Fahrzeuggebrauch von Fahrern in Abhängigkeit von bestimmten Haushaltsmerkmalen und von dem (den) zur Verfügung stehenden Fahrzeugtyp(en) erklären. Sowohl für Haushalte mit nur einem Fahrzeug als auch für solche mit mehreren Fahrzeugen werden verschiedene Modelle auf Basis von Daten zu beobachteten und bekundeten Präferenzen aus einer Umfrage in Kalifornien aus dem Jahr 1993 geschätzt. Die Resultate zeigen, daß die Nutzung mit steigendem Fahrzeugalter abnimmt. Voraussichtlich werden elektrische und andere Fahrzeuge mit eingeschränkter Reichweite weniger genutzt werden als Fahrzeuge mit konventionellen Antriebstechniken. In Haushalten mit mehreren Fahrzeugen werden Nutzungsschwankungen zwischen Fahrzeugen unterschiedlichen Alters, unterschiedlicher Betriebskosten und Art des Brennstoffs ausgemacht. /// Des modèles d’équations des structures sont développés pour expliquer l’affection de conducteur et l’utilisation annuelle des véhicules en fonction des caractéristiques du foyer et de son ou ses véhicules. Des modèles différents sont utilisés pour les foyers possédant un ou plusieurs véhicules, avec pour base des données sur le “revealed preference” et le “stated preference” d’une étude menée en Californie en 1993. Les résultats montrent que l’utilisation du véhicule diminue avec l’âge, et on prévoit que les véhicules électriques ou à fable rayon d’action seront moins utilisés que les véhicules à carburant conventionnel. Dans les foyers possédant plusieurs véhicules, on note des changements dans l’utilisation des véhicules selon l’âge, les frais d’exploitation et le type de carburant utilisé. /// Se desarrollan modelos de ecuaciones estructurales para explicar la asignación de conductores y el uso anual de vehí culos en función de las caracterí sticas de las familias y su(s) vehí culo(s). Se estiman modelos independientes para familias con un vehí culo y familias con varios vehí culos mediante el uso de datos de preferencia revelada y manifiesta a partir de una encuesta realizada en California en 1993. Los resultados cuantifican cómo el uso decrece con la edad del vehí culo y se predice que los vehí culos eléctricos y de serie limitada se usarán menos que los vehí culos de carburante convencional. En familias con varios vehí culos, se detectan cambios en la utilización de vehí culos de diferentes edades, costes operativos y tipos de carburante.

Suggested Citation
Thomas F. Golob, David S. Bunch and David Brownstone (1997) “A vehicle use forecasting model based on revealed and stated vehicle type choice and utilisation data”, Journal of Transport Economics and Policy, 31(1), pp. 69–92. Available at: http://www.jstor.org/stable/20053720.

working paper

Analysis of Costs, Economies of Scale, and Factor Demand in Bus Transport

Publication Date

June 1, 1982

Author(s)

Yossi Berechman

Abstract

A review of the literature on the issue of the cost structure of bus transport shows that most studies use simplistic analytical constructs which do not allow for analysis of relationships between production cost on the one hand and output and input factor prices on the other. In particular, the demand for factors of production, factor substitution and price elasticities are not investigated. This study uses a two factors translog cost function model, which is subject to very few a priori economic restrictions, to investigate these issues. By using a data base which represents the Israeli bus sector, the empirical results provide a comprehensive description of the cost structure of the sector. These results include scale economies, fixed factor proportions-type production technology; non-linear separability of factors in cost function, and small own price elasticity of demand for labor relative to capital.

Suggested Citation
Joseph Berechman (1982) Analysis of Costs, Economies of Scale, and Factor Demand in Bus Transport. Working Paper UCI-ITS-WP-82-3. Institute of Transportation Studies, Irvine. Available at: https://escholarship.org/uc/item/0rm1h3k2.

conference paper

Pooling transportation network company (TNC) rides to the airport to reduce curbside congestion

Proceedings of the 99th annual meeting of the transportation research board

Publication Date

January 1, 2020
Suggested Citation
Karina Hermawan and Amelia Regan (2020) “Pooling transportation network company (TNC) rides to the airport to reduce curbside congestion”, in Proceedings of the 99th annual meeting of the transportation research board.

research report

Low-Carbon Transportation Incentive Strategies Using Performance Evaluation Tools for Heavy-Duty Trucks and Off-Road Equipment

Abstract

This report examines the market share of cleaner technologies and lower carbon intensive fuel use among heavy-duty vehicles (HDVs) and off-road equipment (ORE) to better understand how regulatory measures and incentive programs have and can affect the market. It also projects the uptake of technology for low-carbon transportation (LCT) and identifies the technical features that could potentially improve and optimize the energy demands of both HDVs and ORE under various operational conditions. The research team was led by Principal Investigator (PI) Professor Stephen Ritchie of the University of California, Irvine’s (UCI) Institute of Transportation Studies (ITS), in collaboration with Professor Scott Samuelsen from UCI’s Advanced Power and Energy Program. Research partners Dr. Bo Liu from the University of California, Los Angeles, (UCLA), Dr. Kanok Boriboonsomsin and Fuad Un-Noor from University of California Riverside (UCR), and Suman Mitra from the University of Arkansas (UARK). In order to identify barriers to uptake of LCT, the research team conducted an analysis of existing market survey and real-world operation data of heavy-duty fleets and ORE participating in incentive programs. More incentive programs exist for HDVs; thus, lessons learned were extrapolated to OREs and that sector’s unique challenges to increasing the market share of clean technology were identified. The project also examined incentive programs as a whole and quantified socioeconomic, environmental, and health impacts as a function of incentive dollars spent on clean technology adoption. Using the market data, as well as inputs from other research projects, this project delivered a tool that forecasts low-carbon transportation technology market penetration between 2020 and 2050 that considers incremental cost, projected availability of low-carbon fuel sources, estimated reduction of criteria pollutants, and GHG emissions. Finally, this project estimated the year in which low-carbon transportation technology solutions reach cost parity or market acceptance relative to conventional technologies without incentive program supports.

Suggested Citation
Craig Rindt, Mingqi Yao, Naila Sharmeen, Andre Tok, Stephen G. Ritchie, Blake Lane, Kate Forrest, Mike MacKinnon, Scott Samuelsen, Bo Liu, Kanok Boriboonsomsin, Abdullah Fuad Un-Noor, Suman Mitra, Farzana Mehzabin Tuli and Vuban Chowdhury (2023) Low-Carbon Transportation Incentive Strategies Using Performance Evaluation Tools for Heavy-Duty Trucks and Off-Road Equipment. Draft Final Report. California Air Resources Board / Institute of Transportation Studies, UC Irvine. Available at: https://ww2.arb.ca.gov/sites/default/files/2024-01/IV.3%20-%2019rd026-dfr-rsc_final_2023_12_20_clean.pdf.

conference paper

Learning Representation for Anomaly Detection of Vehicle Trajectories

2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Publication Date

October 1, 2023

Author(s)

Ruochen Jiao, Juyang Bai, Xiangguo Liu, Takami Sato, Xiaowei Yuan, Qi Alfred Chen, Qi Zhu

Abstract

Predicting the future trajectories of surrounding vehicles based on their history trajectories is a critical task in autonomous driving. However, when small crafted perturbations are introduced to those history trajectories, the resulting anomalous (or adversarial) trajectories can significantly mislead the future trajectory prediction module of the ego vehicle, which may result in unsafe planning and even fatal accidents. Therefore, it is of great importance to detect such anomalous trajectories of the surrounding vehicles for system safety, but few works have addressed this issue. In this work, we propose two novel methods for learning effective and efficient representations for online anomaly detection of vehicle trajectories. Different from general time-series anomaly detection, anomalous vehicle trajectory detection deals with much richer contexts on the road and fewer observable patterns on the anomalous trajectories themselves. To address these challenges, our methods exploit contrastive learning techniques and trajectory semantics to capture the patterns underlying the driving scenarios for effective anomaly detection under supervised and unsupervised settings, respectively. We conduct extensive experiments to demonstrate that our supervised method based on contrastive learning and unsupervised method based on reconstruction with semantic latent space can significantly improve the performance of anomalous trajectory detection in their corresponding settings over various baseline methods. We also demonstrate our methods’ generalization ability to detect unseen patterns of anomalies.

Suggested Citation
Ruochen Jiao, Juyang Bai, Xiangguo Liu, Takami Sato, Xiaowei Yuan, Qi Alfred Chen and Qi Zhu (2023) “Learning Representation for Anomaly Detection of Vehicle Trajectories”, in 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 9699–9706. Available at: 10.1109/IROS55552.2023.10342070.

conference paper

Travelbehaviour.com: Activity approaches to modeling the effects of information technology on personal travel behaviour

Travel behaviour research: The leading edge

Publication Date

January 1, 2001

Author(s)

Suggested Citation
TF Golob (2001) “Travelbehaviour.com: Activity approaches to modeling the effects of information technology on personal travel behaviour”, in . Hensher, D (ed.) Travel behaviour research: The leading edge. PERGAMON-ELSEVIER SCIENCE LTD / Int Assoc Travel Behav & Res, pp. 145–183. Available at: 10.1016/B978-008043924-2/50005-5.