Monday, February 3, 2020

Glance over AWS re:Invent 2019


AWS re:Invent is one of the major learning conferences for global cloud computing community. It is hosted by Amazon Web Services. This event consists of keynote announcements, training, technical sessions, certification opportunities and many more. Last time it took place in Las Vegas, USA in December with over 75 launches and announcements of new services and features. 



Which services or features will be available on AWS?


Amazon EC2 Inf1 instance for Machine Learning


When you have to deploy machine learning application in the cloud, you need to consider about deployment instance’s performance as well as its cost. If the instance offers high performance, you can train your model to a great extent and run large scale machine learning inference applications such as image recognition, natural language processing.





Amazon EC2 Inf1 instance offers high performance, low-latency inferencing and the lowest cost with AWS Inferentia chips. AWS Inferentia chip is the first custom ML chip designed by AWS to supports up to 128 TOPS and FP16, BF16, INT8 data types. And it delivers up to 3X higher throughput and up to 40% lower cost per inference compared to GPU powered G4 instances. 

Amazon EC2 Inf1 instances come in 4 instance sizes on On-Demand, Reserved, and Spot Instances types or as part of a Savings Plan. And it is available in the following AWS Regions.

  • US East (N. Virginia)
  • US West (Oregon)

Amazon Braket for Quantum Computing


Quantum computing is capable to solve complex problems in several areas such as optimization, energy storage, chemical engineering etc. Normally these complex problems are beyond the reach of a classical computer. And also we need potentially different approaches and new skills to design quantum applications.


Amazon Braket provides
  • step-by-step guides, resource library and tutorials to learn
  • single environment to design, test and run quantum algorithms
  • access to a variety of different types of quantum computers
  • fully-managed Jupyter notebooks to explore possible applications, visualize results and optimize quantum algorithms
  • publishes event logs and performance metrics to Amazon CloudWatch



So this fully managed web service makes it easier to solve complex problems by doing experiments and explore them with quantum computing.

AWS Fargate on Amazon EKS


AWS Fargate is a service that provides on-demand, appropriate sized compute capacity for elastic containers. Here, You don’t need to worry about resource provisioning, configuring or scaling. So AWS Fargate has the responsibility to look on them. Previously AWS Fargate is only available on Elastic Container Service. But now you can control which pods start on Fargate and how they run on Amazon EKS cluster.



               

Currently, AWS Fargate on Amazon EKS is only available in the 4 Regions. Fargate on Amazon EKS has the following considerations

  • It is only supported to Application Load Balancer and recommend to use the ALB Ingress Controller to triggers the creation of ALB
  • Privileged containers are not supported
  • It is not recommended stateful application for pods running on Fargate
  • GPU-based instances are not available
  • Daemonsets are not supported which ensures that all (or some) Nodes run a copy of a Pod


Amazon S3 Access Points


Amazon Simple Storage Service (Amazon S3) is an object storage service which is used to store any amount of data with 99.999999999% durability for a range of use cases. These data can be accessed by different applications for different purposes. 


Shared bucket access to different applications can be managed by bucket policy. When the number of applications which connect to the S3 grows, the bucket policy becomes more complex and hard to manage. And also changes made on bucket policy may result in an unexpected impact on another application.


To avoid these Issues, AWS introduced Amazon S3 Access Points available in all regions at no additional cost. Here, each application has individualized access point with names and customized permissions




Every access point is
  • Associated with a single bucket
  • Contains a Network Origin Control to limit storage access from your VPC
  • Contains a unique DNS name
  • Configured with an access policy specific to the application
  • Allows separate management
Amazon S3 Access Points can be applied on new buckets as well as existing buckets through the AWS Management Console, the AWS CLI, API or SDK client.


A quick recap about other launches


AWS re:Invent 2019 ended with over 75 launches and announcements of new services and features. Followings are some of the new services they announced in a different type of services.

Compute


  • AWS Graviton2 Processor - For general purpose, compute-intensive, and memory-intensive workloads
  • AWS Nitro Enclaves - Make additional security by isolating sensitive data within EC2 instance
  • AWS Compute Optimizer - Identify optimal Amazon EC2 instances and EC2 Auto Scaling group for workloads using a ML-powered recommendation engine
  • EC2 Image Builder - Build and maintain secure OS images more easily and quickly 

Storage


  • EBS Direct APIs for Snapshots - Provide a set of APIs to directly read EBS snapshot data


Database and Analytics


  • Amazon RDS Proxy - Database proxy feature for Amazon RDS to make applications more scalable, resilient and secure
  • Amazon Redshift RA3 instances with Managed Storage - Optimize data warehouse costs by paying for compute and storage separately
  • Amazon Redshift Federated Query - Analyze data across data warehouse, data lakes, and operational database

Security


  • Amazon Detective - Analyze, investigate and identify the root cause of security findings and suspicious activities.
  • Amazon Fraud Detector - use to detect online fraud in real-time, at scale using machine learning


Service beyond the region


  • AWS Outposts - Extends AWS infrastructure, APIs, services and tools to virtually any connected customer site for low latency or local data processing application needs.
  • Local Zones - Extend the AWS Cloud to more locations and closer to your end-users

Now you have so many AWS services than previous to build cool stuff and enjoy...!

No comments:

Post a Comment

How to send Slack notification using a Python script?

 In this article,  I am focussing on sending Slack notifications periodically based on the records in the database. Suppose we have to monit...