Project description
Processing-in-memory techniques to accelerate relational databases
Relational databases play a crucial role in managing and analysing complex data sets across various sectors. They consist of records and attributes, enabling detailed data scrutiny and informed decision-making. However, the prevailing data analysis model suffers from lengthy execution times and high energy consumption due to the separation of data processing and storage. The ERC-funded Real-Database-PIM project aims to address this challenge by leveraging processing-in-memory (PIM) techniques to accelerate analytical processing in relational databases. By employing bulk-bitwise PIM and capitalising on memory cell arrays, the goal is to reduce data movement, processing time and energy expenditure. The project’s objective is to develop the memristive memory processing unit (mMPU) to accelerate relational database analysis. The mMPU holds the promise of a computing system that is 10 times faster and a hundred times more energy efficient.
Objective
In the contemporary digital era, relational databases play a pivotal role in business applications, facilitating the management and analysis of intricate data sets in diverse sectors including healthcare, finance, and social media platforms. These databases comprise relations represented as sets of records and attributes, allowing for detailed data scrutiny and insightful business decision-making. However, the prevalent model of data analysis is beleaguered by lengthy execution times and considerable energy consumption due to the separation of data processing and storage, necessitating hefty computational investments.
Addressing this, we propose an innovative approach leveraging processing-in-memory (PIM) techniques, specifically bulk-bitwise PIM, to hasten analytical processing in relational databases. Grounded in emergent nonvolatile memristive memory technologies, this method capitalizes on memory cell arrays for concurrent data processing and result storage, markedly diminishing data movement and consequently, time and energy expenditure. Our endeavor is to craft the memristive memory processing unit (mMPU), originally developed during the PI's ERC StG Real-PIM-System project, adept at accelerating relational database analysis, promising to deliver a computing system that is tenfold faster and a hundred times more energy-efficient at a fraction of the current cost, revolutionizing data analysis and offering substantial savings in both time and financial resources.
Fields of science
Keywords
Programme(s)
- HORIZON.1.1 - European Research Council (ERC) Main Programme
Funding Scheme
HORIZON-ERC-POC - HORIZON ERC Proof of Concept GrantsHost institution
32000 Haifa
Israel