Top re:Invent 2021 announcements for the advertising and marketing technology industry
There were 85+ new service announcements over the course of AWS re:Invent 2021, along with other announcements from partners and solution teams, so to make it easy for advertising and marketing technology customers to see relevant news, here’s a list of the top 10 announcements specific to the industry.
1. AWS Graviton3 processors and new compute instance types for improved price performance at scale
- Meet AWS Graviton3. Amazon Web Services (AWS) announced AWS-designed Graviton2 processors in 2019, which were the fastest processors in the cloud with up to 40% better price performance than comparable X86-based instances. This year, we announced Graviton3, which provide up to 25% better compute performance, up to 2x higher floating-point performance, and up to 2x faster cryptographic workload performance compared to Graviton2.
- The first available instances with Graviton3 processors are Amazon EC2 C7g instances, which offer up to 25% better performance over current generation C6g instances. These instances will be a great match for compute-intensive workloads such as batch processing, media encoding, ad serving, bidding, distributed analytics, and CPU-based machine learning (ML) inferencing.
- AWS announced new storage-optimized I-family instances, including Amazon EC2 Im4gn and Is4gen instances powered by AWS Graviton2 processors, and Amazon EC2 I4i instances with third-generation Intel Xeon Scalable processors and new NVMe-based AWS Nitro SSDs. These instances will help customers running real-time database workloads with Aerospike, Cassandra, and Couchbase where low latency local NVMe storage is required.
- AWS announced two new instances that will help improve ML performance for scaled training and inference:
- Amazon EC2 Trn1 instances (preview) powered by AWS Trainium chips provide the best price performance and the fastest time-to-training required by use cases such as natural language processing (NLP), computer vision, search, recommendation, ranking, and more.
- Amazon EC2 G5g instances are powered by AWS Graviton2 processors and feature NVIDIA T4G Tensor Core GPUs to provide the best price performance in Amazon EC2 for graphics workloads such as Android game streaming and are a cost-effective platform for low-latency machine learning inference. They are ideal for deploying deep learning applications that need access to NVIDIA GPUs and their associated AI libraries.
2. LiveRamp launches new solution on AWS for performing identity translation directly inside your VPC
LiveRamp launched Embedded Transcoder on AWS Marketplace, a solution that empowers brands, technology platforms, data providers, and publishers to deploy LiveRamp identity services into their Amazon Virtual Private Cloud (Amazon VPC) and unlock translation services while maintaining control of their data. The solution translates 100% of LiveRamp’s privacy-first, person-based RampID identifiers from one encoding to another without any data drop off, preserving customers’ data from data loss, protecting data quality, ensuring accuracy to power customers’ marketing workflows and use cases. Because the solution runs within the customer’s Amazon VPC, the solution reduces latency and improves performance throughout all downstream use cases including analytics, direct activation, and measurement while abiding by world-class privacy and security standards. LiveRamp’s AWS Embedded Transcoder deployment invokes Amazon SageMaker’s network isolation mode on the appliance, protecting your VPC and deployed LiveRamp technology. This approach helps customers to bring the full power of LiveRamp identity into their own environments while protecting their own data and allowing LiveRamp to protect its data, too. Learn more.
3. Amazon Ads uses AWS to handle millions of ad requests per second
There are stories about scale, and then there’s this re:Invent 2021 breakout session featuring Amazon Ads, which describes how it uses AWS to handle hundreds of millions of requests per second (trillions of ads per day) within a 120-millisecond latency budget for its ad server and latencies under 20 milliseconds for its ML workflows. In the session, Amazon Ads goes into detail on its cloud journey to build campaign management, budgeting, ad serving, event logging, data lake/warehouse, and a microservice inferencing architecture for real-time ML. As just one example of what it takes to handle a workload of this magnitude, the team needed a caching system capable of handling 500 million requests per second! It’s must-see for industry enthusiasts and engineers. Check out Jeff Barr’s blog post about the session or watch the AWS on Air session live from re:Invent to learn more.
4. Aerospike announces it will compile for Arm-based AWS Graviton instances and launches solution for ad platforms on AWS
Aerospike, a real-time NoSQL data platform that powers ultra-low latency applications with predictable submillisecond performance at petabyte scale commonly used across ad tech, announced its plans to launch a solution for Arm-based AWS Graviton2 instances in early 2022. The early results shared suggested they expect “significant” price performance improvements over running Aerospike on comparable X86-based instances. Just prior to re:Invent, AWS and Aerospike also announced a new industry solution on AWS that helps ad tech customers accelerate deployments and optimize performance of Aerospike using an AWS Quick Start deployment, reference architectures, customer examples, and prescriptive benchmarks for running Aerospike on Amazon Elastic Compute Cloud (Amazon EC2).
5. Announcing six new Amazon SageMaker capabilities for ad intelligence workloads
AWS announced six new capabilities for Amazon SageMaker, which helps data scientists and developers prepare, build, train, and deploy high-quality ML models quickly:
- Amazon SageMaker Serverless Inference offers serverless compute for machine learning inference at scale
- Amazon SageMaker Ground Truth Plus offers a fully managed data labeling service that uses a highly skilled workforce and built-in workflows to deliver high-quality annotated data for training machine learning models faster at lower cost
- Amazon SageMaker Studio now makes data engineering, analytics, and machine learning workflows accessible within a universal notebook through built-in integration with Amazon EMR
- Amazon SageMaker Training Compiler helps customers train deep learning models up to 50% faster by automatically compiling code to make it more efficient
- Amazon SageMaker Inference Recommender automatically suggests the optimal AWS compute instances for running machine learning inference with the best price performance
- Amazon SageMaker Canvas expands access to machine learning by providing business analysts the ability to generate more accurate machine learning predictions using a point-and-click interface—no coding required
6. Best practices for using Amazon SageMaker for scaled advertising intelligence and ad platform workloads
Speaking of Amazon SageMaker, we heard from several advertising and marketing technology customers about their use of SageMaker during re:Invent for predictive analytics and real-time advertising use cases.
- Acxiom shared how it reduced its model inference time from days to hours with Spark on Amazon EMR, and then used Amazon SageMaker for propensity scoring with billions of records and thousands of propensity models. Watch the talk on the re:Invent platform.
- AppsFlyer shared how it uses Amazon SageMaker to develop a predictive analytics solution that uses predictive modeling over anonymous data to accurately forecast mobile user lifetime value (LTV) through iOS SKAdNetwork campaigns without permitting identification of individual users. The product can produce insights based on the first 24–48 hours of user interaction. The solution also protects user privacy in compliance with Apple’s iOS 14 privacy changes for advertising. Watch the talk here on the re:Invent platform.
- Tapjoy shared how it uses Amazon SageMaker for real-time mediation for programmatic advertising with latencies of ~20ms. This session was only available for live attendees, but we’ll be sharing a blog post soon with details.
- Amazon Ads went into detail on its ML use cases in its re:Invent 2021 breakout session, highlighting how the team built a microservice inferencing architecture on top of Amazon Elastic Container Service (Amazon ECS) and AWS App Mesh with specific hardware and software optimizations for real-time and asynchronous models. Listen to the Official AWS Podcast featuring a 15-minute chat with Kun Liu, director of ML at Amazon Ads, about the ML workloads described in the talk.
7. New serverless analytics services
Scaled data processing and warehousing are common workloads across virtually every advertising and marketing technology customer. This year, AWS announced three new serverless options for Amazon Redshift, Amazon MSK, and Amazon EMR, which will help customers analyze data at scale without having to configure, scale, or manage the underlying infrastructure:
- Serverless data warehouse with Amazon Redshift Serverless (preview): The new serverless option for Amazon Redshift now makes it even easier to get insights from data quickly without the need to set up, manage, or scale clusters. Customers currently managing their own Amazon Redshift clusters can easily move them to the new serverless option using the Amazon Redshift console or the application programming interface (API) without making changes to their applications. Learn more.
- Serverless data streaming with Amazon MSK Serverless (preview): Customers running Apache Kafka to capture and analyze real-time data streams from ad logs, IoT devices, website clickstreams, and many other sources where dynamic data is continuously generated can now use Amazon MSK Serverless to build, manage, and scale clusters automatically, so you no longer have to worry about capacity planning or unpredictable workloads. Learn more.
- Serverless big data analytics with Amazon EMR Serverless: With Amazon EMR Serverless, customers simply specify the framework they want to run (for example, Apache Spark, Hive, and Presto), and Amazon EMR Serverless provisions, manages, and scales the compute and memory resources up and down as workload demands change. Learn more.
8. AWS Data Exchange for APIs
AWS Data Exchange for APIs empowers customers to easily find, subscribe to, and use third-party API products from providers on AWS Data Exchange, with AWS-native authentication and governance, consistent API documentation, and supported AWS SDKs to make API calls. Many industry customers are also data providers on AWS Data Exchange, including Foursquare, Experian, and Neustar. With AWS Data Exchange for APIs, data providers can now reach millions of AWS customers that consume API-based data by adding their APIs to the AWS Data Exchange catalog, and more easily manage subscriber authentication, entitlement, and billing.
9. New capabilities for networking-intensive workloads
- AWS launched AWS Direct Connect SiteLink, a new capability of AWS Direct Connect that lets you create connections between your on-premises networks through the AWS global network backbone. With SiteLink, for example, an ad tech firm synchronizing real-time bidding sites could connect data centers through Direct Connect locations without sending traffic through an AWS Region. Traffic flows over SiteLink from one Direct Connect location to another following the shortest possible path. Learn more.
- AWS announced the preview release of AWS Cloud WAN, a new wide area networking (WAN) service that helps you build, manage, and monitor a unified global network that manages traffic running between resources in your cloud and on-premises environments. Read the blog.
- AWS Transit Gateway now supports intra-region peering, giving you the ability to establish peering connections between multiple Transit Gateways in the same AWS Region. With this change, different units in your organization can deploy their own Transit Gateways, and easily interconnect them resulting in less administrative overhead and greater autonomy of operation. Read the blog.
- Data transfer is always an interesting topic for networking-intensive advertising industry workloads such as real-time bidding and ad serving. AWS kicked off re:Invent 2021 with price reductions for data transfer out to the internet and data transfer out to Amazon CloudFront, AWS’s content delivery network, which securely delivers content with low latency and high availability. All AWS customers will benefit from these pricing changes, and millions of customers will see no data transfer charges as a result. Learn more.
10. Three new capabilities that help marketers and media companies reduce heavy lifting to deliver personalized consumer experiences:
- Consolidate consumer profiles using Amazon Connect Customer Profiles Identity Resolution: One of the primary workloads required to deliver personalized experiences is the ability to ingest data about the same consumer from multiple marketing channels and perform identity resolution to unify it into a single customer record. This is particularly relevant for contact center use cases where agents often spend time navigating between multiple similar customer profiles across CRM, marketing, billing, shipping, and ticketing systems at the beginning of each contact. Now, Amazon Connect Customer Profiles offers Identity Resolution designed to automatically detect similar customer profiles by comparing name, email address, phone number, date of birth, and address. For example, two or more profiles with spelling mistakes, such as “John Doe” and “Jhn Doe,” can be detected as belonging to the same customer “John Doe” using clustering and matching ML algorithms.
- Introducing intelligent user segmentation with propensity scoring in Amazon Personalize, helping you to run more effective marketing campaigns: Building audience segments with ML typically requires advanced data science capabilities that marketers and media companies have in limited supply in-house. Now Amazon Personalize offers out-of-the-box segmentation ML that can do the heavy lifting required for audience segmentation for you, and score users based on propensity and shared attributes across products, categories, brands, and more.
- Introducing recommenders for personalized experiences for media and entertainment and retail: Amazon Personalize announced recommenders optimized to deliver personalized experiences for common use cases in media and entertainment and retail. For example, media and entertainment applications can drive greater engagement and retention with personalized recommendations like “Top Picks” for users on the welcome screen and “More Like X” on video detail pages where the context of what a user has watched is critical to discover what to watch next. Retail businesses can to highlight “Best Sellers” and the items “Frequently Bought Together” to empower customers to more easily build their baskets at check-out. It is now faster and easier to deliver high performing personalized user experiences in your applications without any ML expertise required.