(Pictured above: Lockheed Martin’s Rick Ambrose on stage at AWS re:Invent, Nov. 27, showing off his company’s new low-cost antenna that streams data to the AWS Cloud.)
By Lynn Greiner
AWS RE:INVENT — By the end of the first day of its annual conference, re:Invent, in Las Vegas, Amazon Web Services had announced a flurry of new services and instances, as well as an unexpected surprise.
VP of global infrastructure Peter DeSantis led off with three new instance types for Amazon Elastic Compute Cloud (EC2). First up were A1 instances powered by Amazon’s new 64-bit ARM-based Graviton processor (yes, AWS is now designing silicon) which it says will offer up to 45 percent cost savings compared with Amazon EC2 general purpose instances, welcome news to partners with customers experiences some cloud sticker shock. AWS says that the instance is ideal for scale-out workloads, and it expects it will be popular among educators and ARM enthusiasts. Graviton has 16 cores per processor and supports Amazon Linux 2, Red Hat and Ubuntu Linux, as well as container services including Amazon Elastic Container Service (Amazon ECS) and Amazon Elastic Container Service for Kubernetes (EKS).
The new C5n and P3dn instances offer 100 Gbps networking throughput to enable high-performance computing (HPC), machine learning training and data analytics. The C5n, available now, is designed for compute-intensive workloads needing high bandwidth, while the P3dn GPU instance, available next week, is touted for HPC and ML.
To shove the bits through faster, AWS also announced two networking offerings. The Elastic Fabric Adapter (EFA) is an upcoming network adapter for EC2 instances that AWS says delivers the performance of on-prem HPC. It is now in preview, and currently is available on C5 and P3dn instances, with support for more instance types coming in early 2019.
AWS Global Accelerator, available today, uses AWS’ global network backbone to direct traffic from users to application endpoints across AWS regions. Traffic is routed based on the client’s geographic location, application health and customer-configurable routing policies. AWS Global Accelerator also allocates a set of static Anycast IP addresses that are unique per application, so clients need not be reconfigured as the application scales.
AWS continuously monitors application endpoints, and Global Accelerator directs clients only to healthy endpoints.
This week, Amazon is releasing SageMaker Neo, an enhancement to SageMaker that allows customers to train once and deploy anywhere by using the Neo Deep Learning compiler, which optimizes for each display target. AWS also released Neo to open source.
The new lightweight virtualization technology beneath AWS Fargate and Lambda serverless offerings, Firecracker, was formally unveiled and released as open source under the Apache 2.0 license. Firecracker is designed to enable secure multi-tenancy, according to a blog by AWS evangelist Jeff Barr, who detailed a series of security features that minimize the attack surface.
Four new services address the challenges of the Internet of Things (IoT).
The big surprise, presented by CEO Andy Jassy, was …