LAS VEGAS – Builders compose apps like builders compose songs, and they want the same kind of tools and capabilities to assist them. That was the theme of AWS CEO Andy Jassy’s keynote at the AWS re:invent conference here today. Jassy used the comparison to provide a thematic unity to many new services – 22 by his count – that he formally announced over his two and a half plus hours onstage.
“The theme that struck me this year is how builders compose apps is similar to how musicians compose songs,” he said. “Both want the freedom to be free from abusive providers,” which would be labels for musicians, and Oracle, for developers. “Both have convictions about their ideas because they are inventing. Both want information about customers so they can predict what they might like as they create. Both don’t want to wait”
Jassy emphasized that AWS already has meaningfully more compute instances than anyone else, including serving specific use cases whether they be memory constrained, IO constrained, or Big Data optimized, to meet specific developer needs. He then proceeded to announce a plethora of new services.
“Over the last 24 months, lots of people have become interested in Kubernetes,” he said. “However, for customers who want to run it on AWS, there’s a lot of work to do, including configuration, load balancing, and rollover.”
Accordingly, AWS is announcing a new Amazon Elastic Container Service for Kubernetes.
“There are lots of things you guys will like about this,” Jassy told his audience. “It’s the latest version of Kubernetes. The service will also deploy Kubernetes across multiple availability zones so you don’t have a single point of failure. There are also automated upgrades and patches.
“We are really excited about this,” he said. “You now have ECS and EKS to choose from.”
“We have been helping customers use container-based storage for years,” said Jeff Barr, Chief Evangelist at AWS. “AWS customers have been getting great value from ElastiFile and Docker. This takes away the grunge work for them. The customer’s goal is to do computations, not to set up Kubernetes, and we want to get as close to their goal as possible.”
Jassy then introduced AWS Fargate, which makes running containers easier than ever by eliminating the need to manage clusters of servers.
“AWS Fargate lets you run containers without having to manage servers and clusters,” he said. “It manages the underlying infrastructure, deploys tasks for you and scales it out across multiple availability zones – with no servers no clusters and no provisioning. It totally changes the way you run containers. It lets you run them at the task level instead of the server level.”
Jassy also said that Fargate would be a huge asset for integrator partners.
“Now that customers have choices for managed container services without having to worry about servers through Fargate, it opens up new opportunities for system integrator partners,” he said.
AWS Fargate is available for Amazon ECS now and will be available for Amazon EKS in 2018.
Jassy then turned to Larry Ellison and Oracle – always a favorite pincushion at this event – in announcing new capabilities for their Aurora database service, complete with the recycled [but still funny] cartoon posted above, which got a good audience reaction.
“The last 20 years have been uncomfortable for companies with database providers they have had to use,” he said. “Oracle this year doubled the price of their product to run on AWS and Microsoft.” In contrast, he said that Aurora is the fastest growing AWS service ever, in part because it is a tenth of the price of these traditional database services.
The enhancement Jassy announced is AWS Aurora MultiMaster, which extends Aurora to support multiple write master nodes across multiple Availability Zones.
“It now creates multiple reads AND writes, and the multiple zones provides zero downtime from any node failure,” he said. “This becomes the first relational database service to scale out across multiple datacenters.
MultiMaster is now in preview for a single region, with the full multi-region support coming in 2018.
For customers with databases that are not fully occupied, and who have variable usage patterns that would make AWS Aurora proper a poor financial choice because they would be paying for too much downtime, Jassy introduced AWS Aurora Serverless.
“With Aurora Serverless, customers will pay only by the second when the database is being used, and don’t have to provision any database instances themselves,” he said. The database automatically starts, scales, and shuts down based on application workload. Customers just create an endpoint through the AWS Management Console, specify the minimum and maximum capacity needs of their application, and Amazon Aurora handles the rest.
“There’s been a lot of excitement at the event around the entire serverless concept, where you don’t need to launch a compute instance to run,” Barr said. “Customers like it because it scales very well and when nothing runs, you aren’t paying for a server that isn’t doing anything.”
Jassy then introduced enhancements to Amazon DynamoDB, their fully managed, scalable NoSQL database service.
“Relational databases don’t work as well in databases that get very large, and that’s why we built Dynamo DB, which scales to really high levels,” he said. “Now we are announcing the launch of DynamoDB Global Tables, the first fully managed multi-master multiregional database in the world for high performance and globally distributed applications.” It is generally available today.
Also announced was Dynamo DB Backup and Restore.
“It provides on-demand continuous backups, with point-in-time restore up to the second for the last 35 days,” he said. “It can back up hundreds of TB of data with no interruption to your application. This is this something that you won’t find anywhere else.” It’s available today, with the point-in-time restore capability available in early 2018.
Jassy then announced Amazon Neptune, a fully managed graph database, which scales automatically, with no downtime or performance degradation, and is designed to offer greater than 99.99 percent availability and automatically detect and recover from most database failures in less than 30 seconds.
“It’s really fast and scalable and allows you to store billions of relationships and query them with millisecond latency,” he said.
Jassy then turned his attention to new data lake-related services.
“Most analytics want to process only a subset of object data,” he said. “For example, a retailer with 20 stores might store all data from all stores in an object each day. If they want sales for the last 30 days from one store, getting it is time consuming. People wanted to filter the data to spend less doing analytics in a data lake.”
“Amazon S3 Select pulls out only the data you need in objects from S3,” Jassy said. “It dramatically improves the ability to filter the data, which speeds up and lowers costs of access by up to 400 per cent.”
Glacier Select, another new offering, makes AWS Glacier a part of a data lake.
“A challenge with archival solutions like Glacier is that is querying takes a long time, so creates issues of not being able to pull out data when you want,” Jassy said. “Customers have asked us to extend the data lake to Glacier. Glacier Select lets you run queries directly in data stored in Glacier.”
While it seemed at times that AWS announced a new service for every tool or platform out there, it wasn’t quite the case. Nothing was announced, for example, around Blockchain.
“We are watching it carefully but we don’t see a lot of practical use cases, yet,” Jassy said. “We don’t build tech because we think tech is cool. We built it when we see a customer problem and see the service as the best way to solve that problem. With Blockchain, we see other ways right now to solve those problems. We are watching it carefully however.”