Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Launch Supplement

University of Waterloo

$12,500.00

Design Optimization of Distributed Middleware for Large-scale Data-intensive Workloads

EF 2024-25

Résumé
Problems that are computationally expensive or have large memory footprint rely on High Performance Computing (HPC) platforms - a large, distributed memory computing cluster with hundreds to thousands of hosts connected via a high-speed network. Distributed computing is enabled by a crucial piece of middleware - the software substrate that sits between user programs and the distributed platform. A distributed middleware offers a high-level interface for application development as well as a runtime that that executes a distributed program, interfaces with operating system and network services. In recent years, the demand for HPC capabilities across diverse industries has been increasing; the key driver here is adoption of data-centric approach to problem solving and decision making, and the growing scale of data science and machine learning tasks. In the next five years, the global HPC market is projected to surpass $50 billion, up from $35 billion today. The core design philosophy of popular HPC middleware such as the Massage passing Interface (MPI) and GASNet, conceived over two decades ago, is largely motivated by problems that are bottlenecked by the speed of host processors. In modern data-intensive problems, however, often distributed communication between cluster nodes is the bottleneck; many exhibits irregular, data-driven computation patterns - leading to workload imbalance and poor scalability. Existing middleware are not designed to regulate extreme network traffic caused by data-intensive problems; often leading to system collapse. Furthermore, popular middleware have limitations in maximizing performance of individual cluster nodes, e.g., limited scalability with increasing core-count, and inadequate support for multithreading - inconsistent with the modern hardware design trend. The long-term goal is to establish a research direction to offer innovation in scalable, reliable and efficient computing for contemporary and emerging data-intensive problems. The research objectives of this Discovery program are tailored to address design shortcomings of current HPC middleware with respect to efficiently and reliably supporting data-intensive workloads on modern hardware: (a) Investigate communication optimization for problems that exhibit unpredictable messaging patterns. (b) Examine design opportunities for ensuring robustness in presence of message-heavy workloads. (c) Explore the design space of computation models that are congruent with the node architecture of modern distributed hardware. Additionally, the training received under this program will equip participating students with a broad skill set in scalable computing and prepare them to assume research and engineering roles in the academia, and industry alike. This research endeavor would strengthen Canada's competence in scalable computing and data science: and from an economic standpoint, would be a step forward toward becoming a key player in the fast-growing HPC industry.
Programme
Discovery Launch Supplement
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
DGECR-2024-00414
Name nom
Reza, Tahsin
Department département
Electrical and Computer Engineering
Organization id
33
Institution établissement
University of Waterloo
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
12500
Program id
DGECR
Program name en
Discovery Launch Supplement
Program name fr
Supplément Tremplin vers la découverte
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2704
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Distributed and parallel processing
Research subject fr
Traitement réparti et simultané
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Design Optimization of Distributed Middleware for Large-scale Data-intensive Workloads
Keywords
distributed computing; parallel computing; high performance computing; middleware; systems software; computer networks; data systems; graph processing; data mining; machine learning
Application summary
Problems that are computationally expensive or have large memory footprint rely on High Performance Computing (HPC) platforms - a large, distributed memory computing cluster with hundreds to thousands of hosts connected via a high-speed network. Distributed computing is enabled by a crucial piece of middleware - the software substrate that sits between user programs and the distributed platform. A distributed middleware offers a high-level interface for application development as well as a runtime that that executes a distributed program, interfaces with operating system and network services. In recent years, the demand for HPC capabilities across diverse industries has been increasing; the key driver here is adoption of data-centric approach to problem solving and decision making, and the growing scale of data science and machine learning tasks. In the next five years, the global HPC market is projected to surpass $50 billion, up from $35 billion today. The core design philosophy of popular HPC middleware such as the Massage passing Interface (MPI) and GASNet, conceived over two decades ago, is largely motivated by problems that are bottlenecked by the speed of host processors. In modern data-intensive problems, however, often distributed communication between cluster nodes is the bottleneck; many exhibits irregular, data-driven computation patterns - leading to workload imbalance and poor scalability. Existing middleware are not designed to regulate extreme network traffic caused by data-intensive problems; often leading to system collapse. Furthermore, popular middleware have limitations in maximizing performance of individual cluster nodes, e.g., limited scalability with increasing core-count, and inadequate support for multithreading - inconsistent with the modern hardware design trend. The long-term goal is to establish a research direction to offer innovation in scalable, reliable and efficient computing for contemporary and emerging data-intensive problems. The research objectives of this Discovery program are tailored to address design shortcomings of current HPC middleware with respect to efficiently and reliably supporting data-intensive workloads on modern hardware: (a) Investigate communication optimization for problems that exhibit unpredictable messaging patterns. (b) Examine design opportunities for ensuring robustness in presence of message-heavy workloads. (c) Explore the design space of computation models that are congruent with the node architecture of modern distributed hardware. Additionally, the training received under this program will equip participating students with a broad skill set in scalable computing and prepare them to assume research and engineering roles in the academia, and industry alike. This research endeavor would strengthen Canada's competence in scalable computing and data science: and from an economic standpoint, would be a step forward toward becoming a key player in the fast-growing HPC industry.