Sunday, September 29, 2024

Subsequent-gen chips, Amazon Q, and speedy S3

AWS re:Invent, which has been happening from November 27 and runs to December 1, has had its regular plethora of bulletins: a complete of 21 at time of print.

Maybe not surprisingly, given the massive potential impression of generative AI – ChatGPT formally turns one yr outdated at this time – numerous focus has been on the AI facet for AWS’ bulletins, together with a main partnership inked with NVIDIA throughout infrastructure, software program, and companies.

But there was loads extra introduced on the Las Vegas jamboree moreover. Right here, CloudTech rounds up the very best of the remaining:

Subsequent-generation chips

This was the opposite main AI-focused announcement at re:Invent: the launch of two new chips, AWS Graviton4 and AWS Trainium2, for coaching and working AI and machine studying (ML) fashions, amongst different buyer workloads. Graviton4 shapes up in opposition to its predecessor with 30% higher compute efficiency, 50% extra cores and 75% extra reminiscence bandwidth, whereas Trainium2 delivers as much as 4 occasions sooner coaching than earlier than and can be capable of be deployed in EC2 UltraClusters of as much as 100,000 chips.

The EC2 UltraClusters are designed to ‘ship the very best efficiency, most vitality environment friendly AI mannequin coaching infrastructure within the cloud’, as AWS places it. With it, prospects will be capable of practice giant language fashions in ‘a fraction of the time’, in addition to double vitality effectivity.

As ever, AWS provides prospects who’re already utilising these instruments. Databricks, Epic and SAP are among the many firms cited as utilizing the brand new AWS-designed chips.

Zero-ETL integrations

AWS introduced new Amazon Aurora PostgreSQL, Amazon DynamoDB, and Amazon Relational Database Companies (Amazon RDS) for MySQL integrations with Amazon Redshift, AWS’ cloud knowledge warehouse. The zero-ETL integrations – eliminating the necessity to construct ETL (extract, rework, load) knowledge pipelines – make it simpler to attach and analyse transactional knowledge throughout varied relational and non-relational databases in Amazon Redshift.

A easy instance of how zero-ETL features might be seen is in a hypothetical firm which shops transactional knowledge – time of transaction, objects purchased, the place the transaction occurred – in a relational database, however use one other analytics software to analyse knowledge in a non-relational database. To attach all of it up, firms would beforehand need to assemble ETL knowledge pipelines that are a money and time sink.

The newest integrations “construct on AWS’s zero-ETL basis… so prospects can rapidly and simply join all of their knowledge, irrespective of the place it lives,” the corporate mentioned.

Amazon S3 Categorical One Zone

AWS introduced the overall availability of Amazon S3 Categorical One Zone, a brand new storage class purpose-built for patrons’ most frequently-accessed knowledge. Knowledge entry pace is as much as 10 occasions sooner and request prices as much as 50% decrease than customary S3. Corporations may also decide to collocate their Amazon S3 Categorical One Zone knowledge in the identical availability zone as their compute assets.  

Corporations and companions who’re utilizing Amazon S3 Categorical One Zone embrace ChaosSearch, Cloudera, and Pinterest.

Amazon Q

A brand new product, and an fascinating pivot, once more with generative AI at its core. Amazon Q was introduced as a ‘new sort of generative AI-powered assistant’ which might be tailor-made to a buyer’s enterprise. “Prospects can get quick, related solutions to urgent questions, generate content material, and take actions – all knowledgeable by a buyer’s data repositories, code, and enterprise techniques,” AWS added. The service can also help firms constructing on AWS, in addition to firms utilizing AWS functions for enterprise intelligence, contact centres, and provide chain administration.

Prospects cited as early adopters embrace Accenture, BMW and Wunderkind.

Need to be taught extra about cybersecurity and the cloud from trade leaders? Try Cyber Safety & Cloud Expo happening in Amsterdam, California, and London. Discover different upcoming enterprise know-how occasions and webinars powered by TechForge right here.

Related Articles

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Latest Articles