Grant: Discovery Grants Program - Individual
McGill University
$41,000.00
Intertwining Databases and Machine Learning: A machine learning driven platform to support in-database analytics
EF 2024-25
Résumé
As data science emerges as a rapidly evolving field and application, the systems community is asked to develop new data and compute platforms that support its many requirements. In this research program, in particular, we want to support the data science explorative phase where data scientists perform declarative data transformations, run analytics functions, and experiment with a wide variety of machine learning algorithms to see which algorithm best fits their data and problem.
The objective is to develop holistic and efficient solutions for this explorative and interactive phase that can be integrated within data management systems and that perform data exploration near-data whenever possible. Data scientists should be able to work with a single, convenient interface. In our work, ML will play a double role. It is part of the data science tasks and must be properly supported by the platform we want to develop. But we also explore ML as a promising tool to build a better platform.
We will be working on three research axes to achieve our medium--term objectives:
A1. Concurrent data science computations:
In research axis A1, we will build a scheduler that controls the execution of concurrent explorative tasks over the available infrastructure. We will first explore how to schedule concurrent ML jobs across available CPUs and GPUs, considering the heterogeneity of the jobs, fairness, and real-time requirements, and then extend this to a system that also schedules other explorative tasks such as analytics functions and declarative queries.
A2. Distributed Data Science
When data is distributed across various data management systems, a near-data approach needs to support distributed solutions for the various tasks in the explorative phase. Our goal is to find a distributed execution architecture that can handle all tasks appropriately, combined with a user-friendly programming model. Furthermore, we will provide efficient mechanisms to handle the parameter state of a model during distributed learning. Model parameters are continuously updated during learning, which by itself is a data management problem that requires efficient consistency management.
A3. Predicting Run-time Behavior
This research axis is a cross-cutting concern that explores the use of machine learning to support the development of our solutions of axes A1 and A2. In particular, we will explore whether machine learning can be used to predict the run-time behavior of the various tasks, e.g., of a particular ML program or a distributed query. Such information will help A1 provide better scheduling decisions and A2 decide on the best execution strategies.
This research aligns well with the 2nd phase of the Pan-Canadian Artificial Intelligence Strategy, as an important component on the path to AI adoption is the availability of platforms that support data science. Canada is also in urgent need of HQP with expertise in applied ML and big data, and this program trains such HQP.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-03970
Name nom
Kemme, Bettina
Department département
Computer Science
Organization id
61
Institution établissement
McGill University
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
Award amount
41000
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
802
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Information systems and technology
Area of application fr
Systèmes d'information et technologie de l'information
Research subject code
2700
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Information technology
Research subject fr
Technologie de l'information
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
Intertwining Databases and Machine Learning: A machine learning driven platform to support in-database analytics
Keywords
system support for data analytics; machine learning for systems; distributed computation; resource management for machine learning; distributed data science; state management for machine learning; data science pipelines; data analytics on dynamic data
Application summary
As data science emerges as a rapidly evolving field and application, the systems community is asked to develop new data and compute platforms that support its many requirements. In this research program, in particular, we want to support the data science explorative phase where data scientists perform declarative data transformations, run analytics functions, and experiment with a wide variety of machine learning algorithms to see which algorithm best fits their data and problem.
The objective is to develop holistic and efficient solutions for this explorative and interactive phase that can be integrated within data management systems and that perform data exploration near-data whenever possible. Data scientists should be able to work with a single, convenient interface. In our work, ML will play a double role. It is part of the data science tasks and must be properly supported by the platform we want to develop. But we also explore ML as a promising tool to build a better platform.
We will be working on three research axes to achieve our medium--term objectives:
A1. Concurrent data science computations:
In research axis A1, we will build a scheduler that controls the execution of concurrent explorative tasks over the available infrastructure. We will first explore how to schedule concurrent ML jobs across available CPUs and GPUs, considering the heterogeneity of the jobs, fairness, and real-time requirements, and then extend this to a system that also schedules other explorative tasks such as analytics functions and declarative queries.
A2. Distributed Data Science
When data is distributed across various data management systems, a near-data approach needs to support distributed solutions for the various tasks in the explorative phase. Our goal is to find a distributed execution architecture that can handle all tasks appropriately, combined with a user-friendly programming model. Furthermore, we will provide efficient mechanisms to handle the parameter state of a model during distributed learning. Model parameters are continuously updated during learning, which by itself is a data management problem that requires efficient consistency management.
A3. Predicting Run-time Behavior
This research axis is a cross-cutting concern that explores the use of machine learning to support the development of our solutions of axes A1 and A2. In particular, we will explore whether machine learning can be used to predict the run-time behavior of the various tasks, e.g., of a particular ML program or a distributed query. Such information will help A1 provide better scheduling decisions and A2 decide on the best execution strategies.
This research aligns well with the 2nd phase of the Pan-Canadian Artificial Intelligence Strategy, as an important component on the path to AI adoption is the availability of platforms that support data science. Canada is also in urgent need of HQP with expertise in applied ML and big data, and this program trains such HQP.