Tag: Amazon S3

Following its rare appearance at last year's event, AWS will preview its fifth-generation self-developed processor, "Graviton 5," built on TSMC's 3nm process, at this year's re:Invent 2025. Apple will again endorse the processor, explaining that it will be used to more efficiently handle the computing needs of its services. On the storage side, Amazon S3 Vectors will be officially launched. These two updates significantly enhance the "brain" of cloud computing and the "memory" of AI applications, respectively, meeting the extreme demands of the generative AI era for computing power and data retrieval. ▲AWS will preview its fifth-generation self-developed processor, "Graviton 5," built on TSMC's 3nm process. Graviton 5: Apple again makes a rare appearance, supporting the operation of iCloud services for 1 billion users. As AWS's most powerful custom chip to date, the Graviton 5 has up to 192 cores per chip. Amazon EC2 M9g instances equipped with this chip offer a 25% improvement in computing performance and better energy efficiency compared to the previous generation Graviton 4. The biggest highlight of the presentation was undoubtedly the appearance of Payam Mirrashidi, Apple's Vice President of Cloud Systems and Platforms, who spoke about the platform. This marks Apple's second endorsement of AWS, following its appearance last year. He revealed that Apple is using Graviton 5's computing power to support its service ecosystem, which handles over 1 billion daily accesses, including iCloud, Swift coding development, and even applications like SMS filtering. Payam Mirrashidi emphasized that implementing Graviton 5 effectively reduced the size of the infrastructure and helped Apple achieve its carbon reduction goals through its lower power consumption. Besides Apple, companies including Airbnb, SAP (which improved HANA Cloud query performance by 35-60%), and Synopsys have also implemented Graviton 5 and achieved significant performance gains. ▲Including Airbnb, SAP (improving HANA Cloud query performance by 35-60%), and Synopsys, among others, have already adopted Graviton5 and achieved significant performance gains. Amazon S3 Vectors Official Release: TwelveLabs Demonstrates 100ms Low Latency. In terms of storage solutions, AWS has officially launched Amazon S3 Vectors, allowing developers to directly store and query vector data in the native S3 environment, significantly simplifying the architectural complexity of AI applications (such as RAG retrieval enhancement generation). ...

At re:Invent 2025, AWS announced a series of major updates to its core storage service, Amazon S3, for the AI ​​and big data era. Among these updates, the official release of Amazon S3 Vectors, supporting native vector storage and querying, was announced, and the maximum storage size of a single Amazon S3 object was increased to 50TB. This helps enterprises build generative AI applications and data lakes more efficiently and cost-effectively. Amazon S3 Vectors Officially Launched: Capable of 2 Billion Vector Indexes, Costs Reduced by 90% To enable AI systems to be stored directly in the native Amazon S3 environment and to perform natural language searches by querying vector data, AWS announced the official release of Amazon S3 Vectors. Compared to the previously released preview version, the official version represents a significant leap in scale, including a single index that can scale to 20 billion vector data points (a 40x increase in capacity), and a single bucket supporting up to 20 trillion vector data points. AWS emphasizes that Amazon S3 Vectors significantly improves performance by 20-3 times for frequent queries, helping customers reduce costs by up to 90% compared to other alternatives and eliminating the complexity of managing dedicated vector infrastructure. This feature also deeply integrates with Amazon Bedrock Knowledge Bases and Amazon OpenSearch Service, allowing enterprises to easily build AI agents and RAG...

Mai-Lan Tomsen Bukovec, AWS Vice President responsible for storage and cloud data migration technologies related to Amazon S3, emphasized how Amazon S3 storage service updates presented at re:Invent 2024 help users drive AI technology innovation from the storage side. ▲Simplifying Cloud Storage Challenges Through Amazon S3 Iterative Updates Since storage plays a crucial role in cloud services, efficient data access has become key to service development. Especially with the current trend of AI technology applications, how to enable AI services to quickly find the correct data and generate correct solutions is closely related to how data is stored. In AI training, training AI models with high-quality data is also closely related to storage, which is why Amazon S3 continues to be an important AWS service. However, with the continuous growth of various services, the amount of data generated daily is also increasing exponentially, forcing enterprises and developers to carefully consider storage architecture planning. Especially with the unimaginable volume of data generated by AI technology development, efficient data access and effective use of this massive data for AI development have become even more critical. Therefore, at re:Invent 2024, AWS touted its ability to help users save over $40 billion in storage costs through the Amazon S3 Intelligent-Tiering storage category. It also added support for the Apache Iceberg open-source storage table format and the Amazon S3 Tables system, which can improve storage speed by up to 3x. Furthermore, the Amazon S3 Metadata management system allows for faster identification of stored data through metadata comparison, ensuring accurate retrieval even if data has changed later. Other updates include addressing the data processing needs of services serving global regions by proposing satellite correction for errors caused by different time zones, ensuring all data accurately reflects its storage and modification time, thus enabling different processing decisions. It also utilizes Aamzon DynamoDB global tables, providing a fully managed, serverless, multi-regional, multi-purpose database with 99.999% availability. Additionally, it introduced the faster distributed SQL database system, Aamzon Aurora SQL. Unstructured data is crucial for the development of artificial intelligence. Mai-Lan Tomsen...