SBIR Phase I: Ultra-High Speed In-Memory Searchable Dynamic Random Access Memory

SBIR 第一阶段：超高速内存中可搜索动态随机存取存储器

• 批准号: 1621443
• 负责人: Wolfgang Hokenmaier
• 金额: $ 22.5万
• 依托单位: Green Mountain Semiconductor, Inc.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-07-01 至 2017-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1621443&HistoricalAwards=false
• 关键词: SBIR; Phase; Ultra; Speed; Memory

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is provided by a novel data processing architecture, utilizing high-parallel in-memory computing for certain recurring and data intensive functions. Traditional computer architecture funnels all data through the CPU. Multiple CPU cores and very high clock frequencies are often used to address the issue of ever increasing demands on data processing capability. This is not sustainable from both an economical and an environmental standpoint, due to the power consumption involved and the lack of performance resulting from the limited parallelism of computation. Datacenters and cloud computing have long lost the ?green label? originally associated with internet technology and software. Their operators need to find more economical ways to process large amounts of data. Linear searches and data indexing can be performed economically within the memory itself, and at a several orders of magnitude increase in performance, due to the inherent parallelism of memory architecture. At the same time, this comes with significant power savings due to the elimination of data transport and CPU clock cycles involved. Datacenters will be able to serve more clients, with fever servers and less power consumption.This Small Business Innovation Research (SBIR) Phase I project investigates the feasibility of in-memory data search and compare algorithms which utilize the inherent repetitive structures of high density memory chips. Initial calculations suggest several orders of magnitude improvement in data throughput by in-place computation and simultaneous elimination of data transport. This addresses the industry's significant need for lower power solutions for so called "Big Data" applications, which are expected to continue their exponential rise through the Internet of Things (IoT), where cloud-stored sensor data are expected to eclipse data uploaded by humans in the near future. This Phase I project will perform a detailed architecture study along with a commercially viable and backward compatible communication protocol as well as behavioral models aimed at establishing the commercial viability and at attracting potential clients and licensing partners. Detailed simulations of the novel digital to analog sensing circuitry will thoroughly investigate the key intellectual property, laying the groundwork for the commercial realization of this innovation.

这个小企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是由一个新颖的数据处理架构提供的，该架构利用高并行内存计算来实现某些重复和数据密集型功能。传统的计算机体系结构将所有数据都通过CPU传输。多个CPU内核和非常高的时钟频率通常用于解决对数据处理能力不断增长的需求的问题。从经济和环境的角度来看，这是不可持续的，因为涉及到的功耗和由于计算的有限并行性而导致的性能缺乏。数据中心和云计算早已失去了？绿色标签吗?最初与互联网技术和软件有关。运营商需要找到更经济的方式来处理大量数据。线性搜索和数据索引可以在内存本身内经济地执行，并且由于内存体系结构固有的并行性，性能可以提高几个数量级。同时，由于消除了涉及的数据传输和CPU时钟周期，这带来了显著的功耗节省。数据中心将能够为更多的客户提供服务，使用更少的服务器和更低的功耗。这个小企业创新研究（SBIR）第一阶段项目调查了内存中数据搜索的可行性，并比较了利用高密度存储芯片固有重复结构的算法。初步计算表明，通过就地计算和同时消除数据传输，数据吞吐量提高了几个数量级。这解决了行业对所谓“大数据”应用的低功耗解决方案的巨大需求，预计这些应用将通过物联网（IoT）继续呈指数级增长，云存储的传感器数据预计将在不久的将来超越人类上传的数据。这个第一阶段的项目将进行详细的架构研究，以及商业上可行和向后兼容的通信协议，以及旨在建立商业可行性和吸引潜在客户和许可合作伙伴的行为模型。对新型数模传感电路的详细仿真将深入研究关键知识产权，为这一创新的商业实现奠定基础。