May 2025 Summaries

The text compares Redis 8.0 with Valkey 8.1, highlighting key differences in their architectures, performance characteristics, and licensing models. Both platforms have made improvements to their I/O threading implementations, but Valkey's dual-channel replication scheme demonstrates more significant benefits for synchronization and write-heavy workloads. Despite these advancements, both systems share limitations, including fork overhead, memory scaling challenges, latency sensitivity, and tuning dependency. Redis 8.0 excels in AI-driven and advanced analytics workloads with its integrated modules and native vector sets, while Valkey is particularly well-suited for traditional caching, queueing, and high-throughput scenarios due to its fundamental optimizations. The choice between the two ultimately depends on individual priorities, with Redis offering advanced features at a cost of complex licensing, and Valkey providing simplicity and efficiency through its BSD 3-Clause License.

Feature stores have emerged as a critical component for managing, serving, and reusing features across different models and teams in the world of machine learning (ML). A feature store is a centralized repository that standardizes the storage, retrieval, and sharing of features. This helps maintain consistency between training and inference, reduces redundant computation, and enables collaboration across teams. Popular feature store options include Feast, Feathr, Hopsworks, AWS SageMaker, and GCP Vertex AI, each with its strengths and considerations for different scale, cost, and performance requirements. The architecture of a feature store consists of several key components, including the offline store, online store, registry, and server, which work together to provide a centralized management system. The dual-layer design of offline and online stores is critical in resolving fundamental tensions in production ML systems, where training requires reproducible historical features with time-travel capabilities, while serving demands sub-10ms access to the latest values. Evaluating offline and online store options involves considering factors such as scale, latency, and operational efficiency, with databases like BigQuery, DuckDB, ScyllaDB, and Dragonfly offering unique strengths for different use cases. In practice, feature stores unlock the ability to build real-time ML applications efficiently, powering systems like real-time bidding and fraud detection, and enabling personalized recommendations at scale.

Dragonfly Cloud is now available on Microsoft Azure, offering a modern and high-performance alternative to traditional Redis-based solutions. Built from the ground up for heavy data workloads in cloud environments, Dragonfly Cloud delivers unmatched performance, scalability, and cost efficiency. It provides a superior alternative to Redis Cloud and Azure Cache for Redis, with better performance and scalability at up to 80% lower infrastructure costs. The platform seamlessly integrates with existing Redis environments, requiring no code changes for migration. It offers a cost-effective and scalable solution that can effortlessly scale to handle workloads of all sizes, reducing the need to overprovision hardware resources. Azure customers can sign up today and have a data store up and running in minutes, with trial credit and waived network fees for 3 months.

Sharp App, a sports betting platform, was struggling with latency issues due to its Redis setup. The team evaluated various alternatives and chose Dragonfly, which significantly reduced latency by 85% and boosted throughput. Dragonfly's vertical scaling capability allowed Sharp App to consolidate workloads onto a single high-performance instance, eliminating the need for clustering. The migration process proved smooth, taking under a day to complete, with minimal configuration adjustments required. After switching to Dragonfly, Sharp App achieved lower overall latency, higher throughput, and simplified infrastructure, ensuring users never miss a profitable opportunity.

Redis 8.0 introduces vector sets and returns to open source with a tri-license model, including AGPLv3, in an effort to rebuild trust after its previous licensing pivot. The company's decision to adopt a new license has been met with skepticism by the community, who have lost trust due to the controversial SSPL licensing change in 2024. Despite this, Redis is trying to win back developer goodwill and Linux distribution inclusion by adding AGPLv3 as an additional option. The open-source world operates on trust, and Redis's decision has burned bridges with the community that had nurtured the project for over a decade. However, competition forces innovation, and Redis's licensing strategy has unwittingly spawned competitors like Valkey, which is now outpacing Redis in certain performance benchmarks. Redis 8.0 also bundles Redis Stack data types and features into a single distribution, simplifying deployment and strengthening its position against Valkey. The new offering includes vector sets, the first new core data type in years, as well as performance improvements, including up to 87% reduction in command latency and 2x more ops/sec throughput. Overall, Redis 8.0 marks a significant shift for the company, with a return to open-source values and a focus on technical innovation.