Redis Announces CEO Transition

Machine learning feature stores

Build scalable low-latency online feature stores for Machine Learning with Redis Enterprise

Leading companies power real-time ML feature serving with Redis online store

As real-time AI/ML based use cases are infused in every facet of our lives, it’s becoming increasingly challenging to deliver ML-based use cases using real-time data with low latency. Built to handle high throughput, low-latency scoring for large datasets, Redis Enterprise is the answer for scalable and affordable online feature store that enables real-time feature serving at scale.

Real-time ML based applications add unprecedented complexity to online scoring

A large and growing portion of ML use cases rely on their online feature stores for online prediction serving while at the same time consuming fresh features. Serving these features with low latency consistently and reliably is very challenging. Doing this at scale is a challenge that traditional databases cannot meet.

Delivering online predictions with fresh data is very challenging

ML-based use cases delivered in real-time using fresh live data directly impacts customer experience or improves business outcomes. However, reliably delivering these predictions online as the user interacts with the app, while at the same time consuming real-time features from streaming sources is very challenging.

Delivering online predictions with fresh data is very challenging
Dataset growth further increases complexity

Dataset growth further increases complexity

With the rise of digital transformation, ML-based applications rely on huge and ever growing datasets with hundreds to thousands of features feeding ML systems at a massive scale. This adds more complexity and cost to the task of consistently and affordably serving features in real-time.

How Redis Enterprise meets the challenges of machine learning online feature stores

machine learning online feature stores diagram

Today’s ML powers mission-critical business use cases like fraud detection and recommendation systems. These applications require reliable and consistent low latency and high throughput serving that can scale potentially to terabytes sized datasets. Your organization’s MLOps platform needs an online feature store that can meet these stringent demands despite the increased complexity, volume and speed.

Serve features in real-time with low latency and high throughput

Serve features in real-time with low latency and high throughput

Get feature lookups from the database for online inference with sub millisecond response latency, keeping up with instant transactions or real-time applications and ensuring great customer experience.

Ensure enterprise grade resilience with five-nines availability

Ensure enterprise grade resilience with five-nines availability

Redis Enterprise offers built-in durability and single-digit-seconds failover with active-active deployment, ensuring zero data loss with no service disruption, suited for your most mission critical flows and data.

Reduce cost at scale without compromising performance

Reduce cost at scale without compromising performance

Redis Enterprise offers multi-tenancy and intelligent tiered access to memory with Redis on Flash, reducing up to 80% of the cost and without compromising performance.

Featured Customers

Product features

Scale efficiently without compromising performance

Linear scaling with sub-millisecond latency

Efficiently scaling database performance is critical for online feature stores. Redis Enterprise scales linearly and with zero downtime to provide more resource-efficient databases that reliably deliver high throughput and sub-millisecond latency.

Intelligent tiered access to memory (DRAM, SSD, persistent memory)

Redis Enterprise offers a cost-effective solution for hosting large datasets by combining DRAM, SSD (Flash), and persistent memory (such as Intel® Optane™ DC). Using an innovative tiered approach that places frequently accessed hot data in memory and colder values in Flash or persistent memory, Redis on Flash delivers high performance similar to Redis on DRAM, while saving you up to 80% on infrastructure costs.

Safeguard your distributed data

Fault tolerance, resilience, and high availability

Redis Enterprise uses a shared-nothing cluster architecture and is fault tolerant at all levels—with automated failover at the process level, for individual nodes, and even across infrastructure availability zones, as well as tunable persistence and disaster recovery.

Enterprise-grade security and compliance

Redis Enterprise ensures production data is isolated from administrative access and offers multi-layer security for access-control, authentication, authorization, and encryption (including data in transit and data at rest).

Active-Active Geo-Distribution

Leveraging Redis Enterprise’s Active-Active database replication with conflict-free replicated data types (CRDTs) enables Feature Stores to gracefully handle simultaneous updates from multiple geographic locations, enabling global scaling of Machine Learning based use cases and applications without compromising latency or availability.

Reduce operational complexity and lower costs

Multiple data types, models and structures

Different data types as string, list, set, hashes, bitmaps, HyperLogLogs and tensors; as well as Redis modules, such as RediSearch, RedisGraph, RedisBloom, and others can be readily applied to use cases like fraud detection, personalization, transaction scoring, and more—all on one integrated data platform reducing operational overhead. By combining multiple Redis modules and data structures, Redis Enterprise can power multiple components of your machine learning platform. The result is a simpler architecture that can process data across multiple models without needing to run multiple database clients and connectors.

Cloud provider and platform integrations

Redis Enterprise is available on all of the major cloud providers as a managed service or as software, provides automation and support for common operational tasks, and integrates with leading machine learning Feature Stores as well as with the platforms underpinning modern software architectures, such as containers and Kubernetes.