Petya Petrova - Fotolia
Redis Labs is continuing to move beyond its caching roots as it builds out a broader set of data features.
The open source database vendor previewed the new technology at the vendor's RedisConf virtual conference on April 20-21.
The foundation of Redis is an open source caching server that has expanded in recent years to become a full database layer, which can enable advanced real-time data services, as well as provide artificial intelligence with RedisAI.
At RedisConf 2021, Redis Labs introduced a series of upcoming capabilities that it said will improve data consistency, search and availability. Redis has found customer traction in recent years, which has led to the vendor, based in Mountain View, Calif., raising new capital in a pair of funding events, including a $100 million round in August 2020 and another $110 million on April 7.
Users outline Redis database use cases
During the conference, users outlined different use cases for Redis, including financial services vendor Capital One.
As part of the April 20 keynote, Mike Lee, head of enterprise payment architecture at Capital One, noted that the company is a big user of Redis. Lee said Capital One processes more than 1.5 billion payments a year and is always looking at improving customer experience by processing payments faster.
Redis Enterprise is a core component of Capital One's platform architecture, helping to strengthen payment visibility for the company, Lee said.
"We can see exactly where a payment is in the value stream at any given point in time," Lee said.
Building world with Redis at Roblox
Online game and game creation platform vendor Roblox is also a Redis user. In a user breakout session on April 21, Jan Berktold, principal engineer at Roblox, outlined how the vendor has evolved of the way it uses Redis over the last two years.
Roblox uses Redis for several purposes, including application caching and for a user notification system.
In 2019, before the COVID-19 pandemic, Roblox was using what Berktold referred to as a "naive" mode that wasn't fully clustered.
Jan BerktoldPrincipal engineer, Roblox
In that approach Redis was serving 5 million requests per second (RPS) for application caching. As usage and demand grew, Roblox adopted the Redis Cluster deployment model that provides a more resilient and scalable approach to Redis with a set of nodes that support each other.
Scalability and performance with the Redis Cluster approach has worked well for Roblox, Berktold said, noting that Roblox now has 80 Redis Clusters in production service, supporting up to 9 million RPS.
"One of the things that surprised me a lot was how far you can actually push nonclustered Redis instances," Berktold said. "Nonclustered Redis instances are simple to operate and can get you far so use them as long as possible and after that Redis clusters are the natural next step."
Ride-sharing giant Uber is another big Redis user. In a session on April 21, Anders Persson, senior software engineer at Uber, said the company is running 20,000 Redis container in production to help provide a data caching layer.
"Redis has a long history at Uber, but it has been used almost exclusively for caching," Persson said.
New Redis real-time data capabilities are coming
During the opening keynote on April 20, Redis Labs founder and CTO Yiftach Shoolman emphasized that in recent years Redis has steadily expanded beyond caching to provide other needed data features, including high availability, clustering and security.
For availability, Redis is building out a feature that Shoolman referred to as 'active-active' Redis.
Shoolman explained that what makes active-active Redis possible is a technology approach known as Conflict-free Replicated Data Types. As part of the expanded Redis offering expected in the second half of 2021, Shoolman said it will be possible to run Redis in an active-active approach that will improve response time for different data types, including JSON.
"Active-active Redis lets you deploy a distributed replicated Redis globally, while always permitting the local speed of read and writes providing the fastest response time," Shoolman said.