At its AWS meeting in San Francisco, Amazon’s cloud computing arm announced today a number of products have been released, including two focusing on non-server files. The first is the GA discovery Amazon Aurora Serverless V2, its server-free data collection service, which can now scale up and down faster than the previous version and be able to measure better growth. The other is GA launches SageMaker Serverless Inference. These two services were first introduced on AWS re: Invent last December.
Swami Sivasubramanian, VP of data, analysis and ML for AWS, told me that more than 100,000 AWS customers are working on their Aurora data center today and that the service will remain the fastest growing AWS service. He noted that initially, type 1, the data storage capacity would take five to forty seconds, and customers would have to double the capacity.
“Because it is server-free, customers then did not have to worry about data capacity management,” Sivasubramian explained. “However, to create a variety of productive workloads [Aurora] “Serverless V1, when we were talking to a lot of customers, they said, customers need the ability to measure fractions in seconds and then a lot more and better, not just double the capacity.”
Sivasubramanian argues that this new system can save users up to 90 percent of data cost compared to the previous power supply cost. He noted that there is no exchange for v2 and that all features in v1 are still available. The team has replaced the global computing machine and storage engine, however, so that it is now possible to estimate this small increase very quickly. “It’s really an amazing technique that the team did,” he said.
Already, AWS customers such as Venmo, Pagely and Zendesk are using this new system, which went into introduction last December. AWS argues that it is not a very heavy lift to replace the workloads currently running on Amazon Aurora Serverless v1 to v2.
Like SageMaker Serverless Inference, which is now also widely available, the Sivasubramanian noted that the service provides businesses with a payment service for importing machine learning models – and especially those that are often sitting – production. Since then, AWS now offers four optional options: No Server Detection, No-Time-Notification of workload while low latency is important, SageMaker Batch Switch for working with batches, and SageMaker Asynchronous Inference for workloads with large load sizes that may require longer processing times. With so many options, it might come as no surprise that AWS also offers it SageMaker Proposal Proposal to help users know the best way to deploy their models.