Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Research Tools and Instruments

University of Toronto

$150,000.00

Infiniband-connected Servers for In-Memory Rack-Scale Computing Research

FY 2015-16

Summary
OBJECTIVES <span style=""line-height: 1.5em;"">Organizations from a wide spectrum of industry sectors rely on data storage and computation in data centres, making the performance, reliability and energy consumption of data centres critical concerns.  Today, data centre computing is being transformed by the commodification of new hardware technologies such as Infiniband networking with RDMA capabilities, specialized compute accelerators, and high capacity SSD storage.  This hardware revolution has profound implications for the design of systems-level software, including storage services such as file systems and caching layers, virtualization services, programming language runtimes, and data processing frameworks.  Traditional performance bottlenecks are being eliminated and new ones are emerging.  For example, the software overhead of traditional storage stacks has become a major problem now that storage hardware can deliver low-latency random access.  Similarly, the software cost of moving data objects from a managed language runtime (e.g., Java) across machines can exceed the network transmission time itself.  Our group is researching these issues, as well as challenges with providing fault tolerance, diagnosing performance problems, and managing virtualized environments. <span style=""line-height: 1.5em;"">REQUESTED EQUIPMENT We are requesting a 16-node server cluster with large main memory capacity to enable in-memory computing frameworks, local SSD drives for tiered storage and caching research, and GPU compute accelerators to study virtualization support for these devices in datacentres.  We are also requesting 100Gb/s network adapters and a switch to interconnect the nodes in the cluster, creating a state-of-the-art rack-scale computing testbed. NEED FOR REQUESTED EQUIPMENT The team of applicants has a well-established track record in ground-breaking systems research, however our current computing equipment is nearly a decade old and unsuitable for research into systems software for next generation rack-scale computing.  We have multiple projects underway that require the requested equipment to take the next steps, validating early emulation results on real hardware, fueling insights into the problems that will arise in future systems, and driving the development of next-generation in-memory rack-scale computing.  It is impossible to conduct the systems-level research we are pursuing on shared computing infrastructure.  Experimental file systems are unsuitable for storing critical data; fault tolerance experiments require the injection of crashes to test recovery; performance measurements require dedicated hardware to eliminate interference; kernel behaviour analysis tools require privileged OS access, which is disallowed in shared computing environments.
Program
Research Tools and Instruments
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RTI-2016-00484
Name nom
DemkeBrown, Angela
Department département
Computer Science (St. George Campus) - Computer Science (St. George Campus)
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2015
Competition year année de concours
2016
Award amount
150000
Program id
RTI
Program name en
Research Tools and Instruments
Program name fr
Outils et instruments de recherche
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1607
Committee name en
Computer, Mathematical and Statistical Sciences (RTI)
Committee name fr
Sciences informatiques, mathématiques et statistiques (OIR)
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
2720
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Computer systems software
Research subject fr
Logiciel de systèmes informatiques
Partie
1
Nb partie
1
Application title
Infiniband-connected Servers for In-Memory Rack-Scale Computing Research
Keyword
NSERC
Application summary
<p>OBJECTIVES</p> <p><span style=""line-height: 1.5em;"">Organizations from a wide spectrum of industry sectors rely on data storage and computation in data centres, making the performance, reliability and energy consumption of data centres critical concerns.&nbsp; Today, data centre computing is being transformed by the commodification of new hardware technologies such as Infiniband networking with RDMA capabilities, specialized compute accelerators, and high capacity SSD storage.&nbsp; This hardware revolution has profound implications for the design of systems-level software, including storage services such as file systems and caching layers, virtualization services, programming language runtimes, and data processing frameworks.&nbsp; Traditional performance bottlenecks are being eliminated and new ones are emerging.&nbsp; For example, the software overhead of traditional storage stacks has become a major problem now that storage hardware can deliver low-latency random access.&nbsp; Similarly, the software cost of moving data objects from a managed language runtime (e.g., Java) across machines can exceed the network transmission time itself.&nbsp; Our group is researching these issues, as well as challenges with providing fault tolerance, diagnosing performance problems, and managing virtualized environments.</span></p> <p><span style=""line-height: 1.5em;""></span>REQUESTED EQUIPMENT</p> <p>We are requesting a 16-node server cluster with large main memory capacity to enable in-memory computing frameworks, local SSD&nbsp;drives for tiered storage and caching research, and GPU compute accelerators to study virtualization support for these devices in datacentres.&nbsp; We are also requesting 100Gb/s network adapters and a switch to interconnect the nodes in the cluster, creating a state-of-the-art rack-scale computing testbed. </p> <p>NEED FOR REQUESTED EQUIPMENT</p> <p>The team of applicants has a well-established track record in ground-breaking systems research, however our current computing equipment is nearly a decade old and unsuitable for research into systems software for next generation rack-scale computing.&nbsp; We have multiple projects underway that require the requested equipment to take the next steps, validating early emulation results on real hardware, fueling insights into the problems that will arise in future systems, and driving the development of next-generation in-memory rack-scale computing.&nbsp; It is impossible to conduct the systems-level research we are pursuing on shared computing infrastructure.&nbsp; Experimental file systems are unsuitable for storing critical data; fault tolerance experiments require the injection of crashes to test recovery; performance measurements require dedicated hardware to eliminate interference; kernel behaviour analysis tools require privileged OS access, which is disallowed in shared computing environments.&nbsp;</p>