June 2021 Summaries

The Room aims to connect top talent globally through a technology-driven platform, leveraging entity-matching algorithms to quickly find high-quality candidates. The core challenge is a mathematically difficult problem requiring accurate and relevant matching of entities, which was addressed by using vector space embedding and combining it with business logic. Redis' high-performance key retrieval based on nearest neighbour vector lookup significantly improved the computation loop, achieving over 15 times speedup without memory overhead. The implementation utilized Scribble Data's Enrich feature store, which handled integration with various data sources, data quality, and batch processing, among other tasks. The optimization focused on reducing resource intensiveness, optimizing for compute distribution, and addressing key constraints such as low latency and high-performance vector similarity computation.

The use of JSON Web Tokens (JWTs) in web authentication has been widely promoted due to its perceived benefits, but it has several potential dangers and complexities that have not been adequately addressed by marketers and creators. JWTs attempt to eliminate database lookup by storing user information in the session token itself, but this approach introduces security issues, such as token revocation problems, where the server has no easy way to revoke a token once it is issued. This can lead to unintended consequences, such as users continuing to access the system until the token expires, or administrators not being able to immediately demote users to lower permissions. Additionally, JWTs can exceed URL length limits and cause issues with data storage and transmission. A more battle-tested approach is to use a database that provides blazing-fast performance, such as Redis Enterprise, which can serve trillions of requests and provide 99.999% availability. By using Redis as a session store and another database as the primary database, developers can achieve a simpler architecture and faster performance.

The blog post discusses using RedisTimeSeries with Apache Kafka for analyzing time-series data. It explains how to set up a data pipeline that ingests, processes, and queries time-series data using Redis and Kafka. The solution involves collecting data from devices and sending it to an MQTT topic, which is then consumed by a Kafka Connect cluster. The data is then stored in RedisTimeSeries for querying. The post also covers configuring Grafana with the Redis Data Source plugin and provides examples of how to use RedisTimeSeries commands such as TS.ADD, TS.GET, and TS.MRANGE to query the data. Additionally, it discusses optimizing RedisTimeSeries, scalability, and integrating it with other tools like Prometheus and Telegraf.

The Redis team hosted the "Build on Redis" Hackathon at RedisConf 21, where developers explored various possibilities within the Redis portfolio and leveraged its advanced capabilities. The event encouraged participants to build projects using modules such as RediSearch and RedisJSON, resulting in innovative applications that showcased the power of Redis. The winners received $100K in prizes, with their projects demonstrating a wide range of use cases, from real-time data platforms to machine learning-based applications, and even gaming and social networking platforms. The event highlighted the passion and creativity of the Redis community, who put significant time and effort into building useful and interesting apps that amazed judges and showcased the capabilities of Redis.

The wealth and asset management industry relies heavily on portfolios as their foundation. With the advent of modern portfolio theory by Harry Makowitz, professionals in this field have been obsessed with maximizing returns while minimizing risk. However, the landscape has changed dramatically with the entry of millions of retail investors into trading infrastructure. The GameStop stock mania is a prime example of how these new entrants can cause market volatility and pressure on brokerages to provide real-time access to prices for their clients. To tackle this challenge, Redis Enterprise, an in-memory database platform, has the potential to handle millions of operations per second with ease. This blog series will cover various real-time use cases in the financial world, including portfolio management and valuation. The first implementation records the id's of all the lots in the account using a SET, while the second model flattens the LOT entity structure and represents each attribute using keys identified by account ID. Both models have an overall time complexity of O(N+T), but Data Model B runs 4.13 ms faster due to reduced network hops. In conclusion, brokerage apps can provide real-time portfolio updates that perform and scale to handle millions of accounts with the help of Redis Enterprise. This data model and architecture can also be applied to use cases beyond securities, such as crypto and ad exchanges.