构建 AI 未来的实时数据库（由 Aerospike 赞助）

关键字: [Amazon Web Services re:Invent 2023, Aerospike, Real Time Database, Aerospike, Machine Learning, Latency, Transactions Per Second, Fraud Detection]

本文字数: 900, 阅读完需: 4 分钟

视频

如视频不能正常播放，请前往bilibili观看本视频。>> https://www.bilibili.com/video/BV1kH4y1C7MV

导读

AI/ML 应用程序渴望海量数据集，并且这种趋势正在成为主流。这意味着您管理持续变化并对数据永不满足需求的解决方案是不可或缺的。了解领先的组织如何标准化使用高效的云数据库来处理指数级增长。Aerospike首席执行官 Subbu Iyer 与 Riskified 的工程副总裁 Guy Machlev 和 Freewheel（Comcast 子公司）的数据副总裁 Xinyu Huang 将展示基于亚马逊云科技的实时应用程序设计，如何帮助企业应对当今动荡的市场压力。亚马逊云科技专家讨论了亚马逊云科技基础设施如何独特地设计用于支持大规模、高增长、任务关键型的工作负载。本讲座由亚马逊云科技合作伙伴 Aerospike 为您带来。

演讲精华

以下是小编为您整理的本次演讲的精华，共600字，阅读时间大约是3分钟。如果您想进一步了解演讲内容或者观看演讲全文，请观看演讲完整视频或者下面的演讲原文。

The video discusses the practical challenges faced by enterprises in leveraging artificial intelligence (AI) and machine learning (ML) to achieve tangible business value. As speaker Linh Ly Gwenap explains, although AI has the potential to bring about significant improvements in efficiency and cost savings, successful deployment requires massive computational power, data storage, and real-time data pipelines – all of which can be expensive. Gwenap highlights statistics such as the explosive power consumption of computation, particularly for training compute-intensive AI models. Some estimates suggest that training a single AI model may have a carbon footprint comparable to a transcontinental flight across the Americas.

However, Gwenap points out that if used wisely, AI also addresses issues like energy consumption. The key is to build systems that can fully utilize the potential of AI. One major challenge is to consume enough real-time data so that AI models can provide useful, context-related recommendations that align with the current situation of the enterprise and the world at large. Therefore, businesses face the pressure of constantly ingesting new data to keep the models updated. The goal is to have enough background information to understand what is happening rather than just yesterday's events and to quickly leverage AI-generated insights and suggestions. This requires an efficient basic data infrastructure.

Aerospike provides a high-performance real-time database to meet these demands. Amit Shah of Amazon Web Services (亚马逊云科技) and Shay Peretz of fraud prevention platform Riskified discuss how Aerospike's technology helps customers satisfy the speed and scalability requirements of real-time AI applications.

Shah explains that clients like Riskified, an Aerospike customer, process millions of transactions per second with sub-millisecond latency requirements. To support this throughput while maintaining low latency, Amazon Web Services offers a wide range of custom computational, memory, storage, and network features tailored to these workloads. For example, Amazon Web Services has collaborated with Aerospike to optimize its Nitro SSD storage system to specifically cater to Aerospike's database needs. This reduces delays for joint customers by 70% and increases cost-effectiveness by 25% – representing significant gains in the cutting-edge field.

Amazon Web Services (亚马逊云科技) has partnered with Aerospike to enable the core database as well as its latest graph database and VectorDB to run on Amazon Web Services' based Arm Graviton processor. Graviton provides high performance at low cost, which is crucial for real-time AI. This joint innovation allows customers to affordably and sustainably leverage AI.

From a customer's perspective, Peretz explained how Riskified utilizes Aerospike as a reliable and scalable data pipeline for its fraud detection platform. Riskified processes billions of e-commerce transaction data points, enriching them in the Aerospike database before providing them to applications for real-time fraud scoring and AI model training. With Aerospike, Riskified can ingest the large and diverse data required to train highly accurate AI models and drive real-time approvals and rejections of transactions. This prevents fraud while minimizing false positives that could annoy legitimate customers.

By combining Aerospike's high-performance real-time database with Amazon Web Services' cloud infrastructure, storage, and processing, customers like Riskified can deploy their AI models into production to achieve concrete business value. The collaboration between Aerospike and Amazon Web Services supports enterprises in overcoming technical and economic challenges in real-time AI, harnessing these powerful technologies to drive efficiency, growth, and innovation.

In summary, this video highlighted the infrastructure challenges in real-time AI and how Aerospike's database and Amazon Web Services' computing capabilities work together to help customers meet the speed, scalability, sustainability, and cost requirements needed to deploy AI in the real world. This partnership enables businesses to intelligently unlock the potential of AI.

下面是一些演讲现场的精彩瞬间：

领导者们讨论了在现实生活中应用人工智能和机器学习所面临的挑战。

虽然人工智能有潜力解决许多问题，但在我们利用海量数据训练大型模型的过程中，我们必须认识到其能源消耗的问题。

通过亚马逊云科技，如播客之类的公司得以存储海量数据并在实时中进行提取，从而为客户提供个性化的体验。

为了构建一个完整的数据平台，需要将实时数据库集成到具有多个摄入点的整体数据管道中，以支持机器学习应用程序的运行。

Aerospike选择亚马逊云科技作为合作伙伴，正是因为其强大的性能能力，足以支持实时数据库和集成的机器学习功能。

亚马逊云科技提供了高性能的云计算服务，满足了客户在低延迟环境下对快速交易处理的需求。

总结

该视频探讨了Aerospike的实时数据库如何在实际场景中助力企业应用AI和机器学习模型。在数据呈指数级增长的背景下，企业需实时获取和分析数据以保持竞争优势。然而，实时数据处理的计算成本往往较高。

作为一家专注于高性能实时数据库的公司，Aerospike致力于优化吞吐量和快速响应时间，以满足各种严苛需求。例如，Riskified这样的客户每秒可处理数十亿笔交易，延迟低于毫秒，得益于Aerospike的支持。通过与亚马逊云科技的紧密合作，Aerospike充分利用后者的丰富实例类型（如Graviton和Nitro SSD）以实现最佳性能价格比。同时，Aerospike的地域分布式架构支持全球覆盖和业务连续性。

此外，Aerospike的store特性和即将推出的vector DB旨在使其实时数据库成为构建实时AI应用程序的理想平台。通过紧密整合数据获取、模型训练和推理过程，Aerospike助企业充分发挥数据价值。

演讲原文

https://blog.csdn.net/just2gooo/article/details/134817870

想了解更多精彩完整内容吗？立即访问re:Invent 官网中文网站！

2023亚马逊云科技re:Invent全球大会 - 官方网站

点击此处，一键获取亚马逊云科技全球最新产品/服务资讯！

点击此处，一键获取亚马逊云科技中国区最新产品/服务资讯！

即刻注册亚马逊云科技账户，开启云端之旅！

【免费】亚马逊云科技“100 余种核心云服务产品免费试用”

【免费】亚马逊云科技中国区“40 余种核心云服务产品免费试用”

亚马逊云科技是谁？

亚马逊云科技（Amazon Web Services）是全球云计算的开创者和引领者，自 2006 年以来一直以不断创新、技术领先、服务丰富、应用广泛而享誉业界。亚马逊云科技可以支持几乎云上任意工作负载。亚马逊云科技目前提供超过 200 项全功能的服务，涵盖计算、存储、网络、数据库、数据分析、机器人、机器学习与人工智能、物联网、移动、安全、混合云、虚拟现实与增强现实、媒体，以及应用开发、部署与管理等方面；基础设施遍及 31 个地理区域的 99 个可用区，并计划新建 4 个区域和 12 个可用区。全球数百万客户，从初创公司、中小企业，到大型企业和政府机构都信赖亚马逊云科技，通过亚马逊云科技的服务强化其基础设施，提高敏捷性，降低成本，加快创新，提升竞争力，实现业务成长和成功。