This Guidance uses EventBridge, which lets you make small architectural changes and simplifies data redirection. Kinesis Data Streams supports a configurable retention period of its streaming data to reduce downstream impact in the event of downtime. You can also use the lifecycle management features of the Amazon S3 Intelligent-Tiering storage class so that raw ML data can automatically move to the optimal access tier based on access frequency. Additionally, you can use CloudWatch for detailed monitoring and logging.

Read the Operational Excellence whitepaper 

This Guidance uses Amazon Virtual Private Cloud (Amazon VPC) so that your data resides only within a network under your full control. Kinesis Data Streams encrypts your data in transit and also provides server-side encryption to automatically encrypt your data at rest. Amazon S3 encrypts all object uploads to all buckets. You can also block public access to all of your objects at the bucket- or the account-level with the Amazon S3 Block Public Access feature.

Read the Security whitepaper 

This Guidance uses Kinesis Data Streams, which lets you configure a seven-day data retention period so that downstream systems can reprocess data in the event of data loss or a processing failure. Amazon S3, which is designed to provide 99.999999999 percent object durability, stores objects redundantly across multiple facilities to increase the reliability of your data storage. Additionally, Lambda runs functions in multiple Availability Zones (AZs) so that it can complete processing in the event of a service interruption to a single AZ.

Read the Reliability whitepaper

This Guidance uses Lambda, which automatically provisions separate implementation environments for each concurrent request so that it can scale to meet your capacity needs without overprovisioning resources. You can configure the Transmission Control Protocol (TCP) keep alive feature to avoid creating new TCP connections for subsequent function invocations. Additionally, Kinesis Data Streams provides an on-demand capacity mode that automatically scales to accommodate your workload throughput needs.

Read the Performance Efficiency whitepaper 

This Guidance uses Kinesis Data Streams, which provides a managed, serverless architecture for data streaming, so that you don’t need to deploy, configure, or maintain streaming server hardware and software. You only pay for what you use, and you can shift to more cost-effective provisioned capacity when traffic is steadier, reducing costs. Additionally, enhanced shard-level monitoring lets you gain insights into traffic patterns so that you can merge underused shards for further cost savings. This Guidance also lets you use S3 Intelligent-Tiering to automatically move your data to the most cost-effective access tier.

Read the Cost Optimization whitepaper 

This Guidance uses serverless services like Kinesis Data Streams and Lambda, which distribute their environmental impact across many users through multi-tenant control planes. Additionally, Kinesis Data Streams provides an on-demand capacity mode, which uses automatic scaling so that only the resources required to handle the current workload are running. Likewise, Lambda automatically scales the number of implementation environments up and down so that no idle resources are running. ARM-based AWS Graviton processors increase the price performance of Lambda by up to 34 percent over x86-based functions, further minimizing hardware requirements.

Read the Sustainability whitepaper