June 2019 Summaries

Today, Redis has announced the general availability of its new module, RedisTimeSeries v1.0. This module is designed to enhance time series data management with Redis and follows extensive beta testing over six months. The benchmark results show that RedisTimeSeries achieved a performance rate of 125K queries per second compared to other time series approaches in Redis. The new module offers significant improvements in memory consumption, query performance, and ease of use for managing time series data.

Redis Enterprise has set a new industry performance record by delivering over 200 million operations per second (ops/sec) on as little as 40 AWS instances, with under 1 millisecond latency. This represents a 2.6X performance improvement in less than 15 months. The company's shared-nothing architecture allows for linear scalability and efficient utilization of system resources, enabling significant improvements in throughput while keeping latency low. By leveraging the new AWS C5 instance family and proper NUMA configuration, Redis Enterprise achieves over 200M ops/sec with a 40-node cluster, breaking its own database performance record from 15 months ago.

Here's a neutral and interesting summary of the text in one paragraph: When designing an Industrial Internet of Things (IoT) system, several key components must be considered including sensors, networks, data stores, and management systems. Sensors are the "eyes and ears" of the system, and their selection depends on various criteria such as cost, environmental conditions, and performance requirements. There are three main types of sensor storage models: short-term, long-term, and none, each with its own advantages and trade-offs. In IoT environments where high volumes of data storage are required, RedisEdge from Redis can be an excellent complement, offering multi-model database capabilities, blazing-fast performance, and reliability. Ultimately, the choice of sensor depends on analyzing short-term and longer-term application requirements to determine the ideal sensor for a specific use case.

Redis Enterprise is the industry's high-performance database of choice for digital transformation initiatives in finance, simplifying the development of real-time applications and addressing data silos by providing an elegant multi-model database platform that brings together heterogeneous data. It offers sub-millisecond response times, predictable performance, and scaling capabilities to support various use cases such as time-sensitive financial transactions, caching, search capabilities, fraud detection and prevention, session store replication, and personalized experiences. With deployment options including hosted and serverless DBaaS, on-premises installation, containers, and cloud or private data center deployment, Redis Enterprise empowers application developers with great flexibility to build high-performance, reliable, and scalable real-time applications.

At RedisConf 2019, Redis introduced a new module called RedisAI, which aims to bring together machine learning (ML) and deep learning (DL) by executing AI models as close as possible to where data currently resides. This can help reduce DevOps challenges associated with moving large amounts of data between applications. RedisAI provides new data structures and allows Redis to manage requests to run ML/DL models and execute them, reducing network overhead and eliminating the need for serializing/deserializing data. Users can install RedisAI using Docker or by building it from source code. The module comes with an example image classification app that demonstrates its functionality in a practical setting.

The choice of a NoSQL database is crucial for building online and operational applications, as it can significantly impact the quality of the user experience. To select the right database, one must define its specific function in the data pipeline, consider latency requirements, throughput needs, and the type of data being stored. Developers often choose NoSQL databases due to their ability to handle semi-structured or unstructured data with flexible schema, simple query patterns, high-velocity transactions, large volumes of data, and quick scalability via distributed computing and storage. The CAP theorem can be used to prioritize characteristics such as consistency, availability, and partition tolerance, helping determine the best fit for an application. Additionally, managed services solutions and self-hosted options are available, with benefits including reduced maintenance, increased scalability, and improved performance. Ultimately, selecting a well-thought-out NoSQL database choice is essential for delivering instant experiences and meeting user expectations.

Transactions in Redis require a deeper understanding than just storing data. Unlike other databases, Redis is single-threaded and has tunable durability, which affects transactions. It lacks rollbacks but can fail a transaction if a key changes before it starts. Understanding the distinction between syntax errors and semantic errors in Redis is crucial for managing transactions effectively. WATCH command enables developers to observe keys for changes and ensure data consistency during transactions. The MULTI/EXEC pattern helps guard against command mismatches, type mismatches, and other potential issues that may arise during transactions.

The speaker presented on combining Redis + Structured Streaming with Apache Spark for scaling continuous applications, highlighting new features in both technologies and their potential for real-time data analysis. They showcased a demo using a Node.js app to collect user activity data in Redis Streams and sink it to Apache Spark. The audience asked interesting questions, and the speaker emphasized growing interest in continuous processing and data streaming, publishing an article on setting up Redis Streams and Apache Spark with the Spark-Redis library.