Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

University of Waterloo

$31,000.00

Co-designing Datacenter Applications with Next-Generation Memory Systems

FY 2024-25

Summary
Today, the majority of computation takes place in large-scale datacenters. A trend in modern data center applications is the increased demand for memory. A study on Google workloads showed that more than half of the execution time is due to access delays from the memory backends. To satisfy the ever--growing demand for larger memory footprints, there is an increased need for extending the memory capacity while maintaining high performance. Unfortunately, the scaling of memory capacity, i.e., DRAM, is facing a limit as Moore's Law is slowing down (i.e., the density improvement of transistors is slowing down). To meet the ever--growing demand, new memory technologies are being developed. For example, the Compute Express Link (CXL) enables an extension of high--capacity, cache coherent memory to servers. On the other hand, new memory technologies, such as Optane 3D XPoint, Phase Change Memory, and Resistive Memory, push the memory density to a new level. These new and emerging memory technologies have the potential of extending memory, but they bring new challenges to system designs. These memories have different access latency, bandwidth, and capacity characteristics, increasing the heterogeneity of the system. Memory devices with high capacity tend to have a higher memory access cost, e.g., CXL--based memory devices extend a server's memory capacity but are slower than main memory. Thus, on the hardware side, the question is how we can design a memory system that provides high capacity without degrading performance. At the same time, legacy datacenter applications are designed for conventional, homogeneous memories. Directly adapting these applications to a heterogeneous system can lead to a significant degradation in performance. Therefore, on the software side, the research question is how can datacenter applications better leverage the different next--generation memory devices. The long-term goal of this proposal is to design next--generation memory systems with high heterogeneity and new near--data computational capabilities for datacenters. We focus on three main objectives: (1) create a benchmark for datacenter applications and characterize these workloads, (2) design the hardware system for the next-generation memory technologies using insights learned from our characterization, and (3) co-design applications to leverage the heterogeneous memory system. These research directions will be cross--disciplinary, spanning computer architecture, software engineering, and computer systems. We expect that our research will help design future--generation datacenters from both the software and hardware perspectives. The research outcome can benefit the Canadian economy by improving the performance of computing systems, and benefit the environment by improving efficiency and reducing the carbon footprint. This research program will also train high-in-demand HQPs in the area of computer science and engineering.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2023-03478
Name nom
Liu, Sihang
Department département
David R. Cheriton School of Computer Science
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
2023
Award amount
31000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
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
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2702
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Computer systems organization
Research subject fr
Organisation des systèmes informatiques
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
Co-designing Datacenter Applications with Next-Generation Memory Systems
Keywords
computer systems; computer architecture; memory systems; performance/system profiling; software-hardware co-design; heterogenous systems; near-data processing; carbon-footprint-aware computing; programming models; compiler
Application summary
Today, the majority of computation takes place in large-scale datacenters. A trend in modern data center applications is the increased demand for memory. A study on Google workloads showed that more than half of the execution time is due to access delays from the memory backends. To satisfy the ever--growing demand for larger memory footprints, there is an increased need for extending the memory capacity while maintaining high performance. Unfortunately, the scaling of memory capacity, i.e., DRAM, is facing a limit as Moore's Law is slowing down (i.e., the density improvement of transistors is slowing down). To meet the ever--growing demand, new memory technologies are being developed. For example, the Compute Express Link (CXL) enables an extension of high--capacity, cache coherent memory to servers. On the other hand, new memory technologies, such as Optane 3D XPoint, Phase Change Memory, and Resistive Memory, push the memory density to a new level. These new and emerging memory technologies have the potential of extending memory, but they bring new challenges to system designs. These memories have different access latency, bandwidth, and capacity characteristics, increasing the heterogeneity of the system. Memory devices with high capacity tend to have a higher memory access cost, e.g., CXL--based memory devices extend a server's memory capacity but are slower than main memory. Thus, on the hardware side, the question is how we can design a memory system that provides high capacity without degrading performance. At the same time, legacy datacenter applications are designed for conventional, homogeneous memories. Directly adapting these applications to a heterogeneous system can lead to a significant degradation in performance. Therefore, on the software side, the research question is how can datacenter applications better leverage the different next--generation memory devices. The long-term goal of this proposal is to design next--generation memory systems with high heterogeneity and new near--data computational capabilities for datacenters. We focus on three main objectives: (1) create a benchmark for datacenter applications and characterize these workloads, (2) design the hardware system for the next-generation memory technologies using insights learned from our characterization, and (3) co-design applications to leverage the heterogeneous memory system. These research directions will be cross--disciplinary, spanning computer architecture, software engineering, and computer systems. We expect that our research will help design future--generation datacenters from both the software and hardware perspectives. The research outcome can benefit the Canadian economy by improving the performance of computing systems, and benefit the environment by improving efficiency and reducing the carbon footprint. This research program will also train high-in-demand HQPs in the area of computer science and engineering.