memory-centric architecture

Extracted Attributes

• Field

  Computer Hardware Architecture

• Key Benefits

  Reduced latency, lower energy impact, and high parallelism

• Core Mechanism

  Computation capability in or near memory arrays

• Alternative Name

  Processing-in-memory (PIM)

• Primary Objective

  Reducing energy and performance costs of data movement

• Energy Efficiency Metric

  Addresses the >90% of energy spent on data movement in AI models

Timeline

• Publication of 'Memory-Centric Architectures: Why and Perhaps What' introducing the DataScalar and Dynamic Data Threads (DDT) architectures. (Source: University of Texas at Austin)

  1997-01-01

• Submission of 'Memory-Centric Computing: Recent Advances in Processing-in-DRAM' to arXiv, detailing functionally-complete bulk-bitwise operations in DRAM. (Source: arXiv:2412.19275)

  2024-12-26

Database-centric architecture

Database-centric Architecture or data-centric architecture has several distinct meanings, generally relating to software architectures in which databases play a crucial role. Often this description is meant to contrast the design to an alternative approach. For example, the characterization of an architecture as "database-centric" may mean any combination of the following: using a standard, general-purpose relational database management system, as opposed to customized in-memory or file-based data structures and access methods. With the evolution of sophisticated DBMS software, much of which is either free or included with the operating system, application developers have become increasingly reliant on standard database tools, especially for the sake of rapid application development. using dynamic, table-driven logic, as opposed to logic embodied in previously compiled programs. The use of table-driven logic, i.e. behavior that is heavily dictated by the contents of a database, allows programs to be simpler and more flexible. This capability is a central feature of dynamic programming languages. See also control tables for tables that are normally coded and embedded within programs as data structures (i.e. not compiled statements) but could equally be read in from a flat file, database or even retrieved from a spreadsheet. using stored procedures that run on database servers, as opposed to greater reliance on logic running in middle-tier application servers in a multi-tier architecture. The extent to which business logic should be placed at the back-end versus another tier is a subject of ongoing debate. For example, Toon Koppelaars presents a detailed analysis of alternative Oracle-based architectures that vary in the placement of business logic, concluding that a database-centric approach has practical advantages from the standpoint of ease of development and maintainability and performance. using a shared database as the basis for communicating between parallel processes in distributed computing applications, as opposed to direct inter-process communication via message passing functions and message-oriented middleware. A potential benefit of database-centric architecture in distributed applications is that it simplifies the design by utilizing DBMS-provided transaction processing and indexing to achieve a high degree of reliability, performance, and capacity. For example, Base One describes a database-centric distributed computing architecture for grid and cluster computing, and explains how this design provides enhanced security, fault-tolerance, and scalability. an overall enterprise architecture that favors shared data models over allowing each application to have its own, idiosyncratic data model. Even an extreme database-centric architecture called RDBMS-only architecture has been proposed, in which the three classic layers of an application are kept within the RDBMS. This architecture heavily uses the DBPL (Database Programming Language) of the RDBMS. An example of software with this architecture is Oracle Application Express (APEX).

Web Search Results

• [PDF] Memory-Centric Architectures: Why and Perhaps What

  The second memory-centric architecture that we describe here is called DDT, for Dynamic Data Threads. In a DDT machine, the memory is distributed among multiple processors, as with a DataScalar architecture, but computation along a local dependence chain occurs uniquely at one node. When a data dependence spans nodes, the source register value is broadcast to all processors, and the intermediate instructions are squashed at all processors except for the owning processor that executed them. [...] Increasing communication costs will force microprocessor-based systems to be more and more partitioned. We argue that this partitioning must eventually include the system memory, and that for codes that are hard to analyze statically, the problem partitioning will be done dynamically, based on the data decomposition. We call such architectures memory-cen-tric, and describe two (DataScalar and DDT) that are evolutionary first steps in this unconventional direction. [...] dynamically based on the given data decomposition. It is this concept that forms the basis of what we call memory-centric architectures.

• Memory-Centric Computing: Recent Advances in Processing-in-DRAM

  > Abstract:Memory-centric computing aims to enable computation capability in and near all places where data is generated and stored. As such, it can greatly reduce the large negative performance and energy impact of data access and data movement, by 1) fundamentally avoiding data movement, 2) reducing data access latency & energy, and 3) exploiting large parallelism of memory arrays. Many recent studies show that memory-centric computing can largely improve system performance & energy [...] We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate > cs > arXiv:2412.19275 # Computer Science > Hardware Architecture arXiv:2412.19275 (cs) [Submitted on 26 Dec 2024] # Title:Memory-Centric Computing: Recent Advances in Processing-in-DRAM Authors:Onur Mutlu, Ataberk Olgun, Geraldo F. Oliveira, Ismail Emir Yuksel [...] > This work describes several major recent advances in memory-centric computing, specifically in Processing-in-DRAM, a paradigm where the operational characteristics of a DRAM chip are exploited and enhanced to perform computation on data stored in DRAM. Specifically, we describe 1) new techniques that slightly modify DRAM chips to enable both enhanced computation capability and easier programmability, 2) new experimental studies that demonstrate the functionally-complete bulk-bitwise

• The Next Computing Revolution - IEEE Computer Society

  Memory-centric computing (or processing-in-memory) is another fascinating research area where we can change the paradigm of how we do computing. Moving data between memory and processor consumes orders of magnitude more energy than computation. Many results from real systems show that most (e.g., >90%) of the energy spent executing major AI models comes from data movement and memory access, not computation performed on the data. Unfortunately, our existing processor-centric design paradigm is [...] With advancements in memory-centric (i.e., processing-in-memory) architectures, what challenges do you foresee in integrating these systems into mainstream computing? [...] Image 6Image 7 Dr. Onur Mutlu is a renowned computer scientist and Professor at ETH Zurich whose pioneering research in computer architecture, memory systems, and hardware security has shaped industry standards and influenced technologies used by billions worldwide. In this article, he shares his insights on advancing memory-centric computing, exploring the challenges, breakthroughs, and future possibilities of next-generation computer memory systems.

• Why Memory-Centric Architecture Is The Future Of In-Memory ...

  Memory is positioned at the heart of the design to provide faster data access than traditional disk-based systems, resulting in reduced latency

• MEMORY-CENTRIC SYSTEM DESIGN TO ACCELERATE AI ...

  This dissertation examines memory-centric system design as a unifying strategy for addressing these challenges. Across four complementary