eHarmony is an online relationship services provider that makes up to 15 million matches per day. To achieve a responsive and interactive customer service experience with the least operational overhead possible, eHarmony relies on Redis Labs for enterprise-grade Redis. Redis Enterprise was selected for many roles – from real-time user authentication and analytics, to supporting low latency match searches, powering news feeds and profile data.
At eHarmony, trust is paramount
Known as the number one trusted dating site, eHarmony has been matching like-minded singles since the turn of the century, making it one of the most enduring and successful online relationship services in the world. As its more than 54 million user base has grown, along with its feature set and supported platforms, maintaining the high performance and seamless real-time customer experience its users have come to trust, is a top priority for the company.
Matchmaking in real-time isn’t easy
eHarmony uses complex machine learning algorithms to anticipate the compatibility of pairs of users. Unlike one-directional recommendations such as those offered by Netflix or Amazon, eHarmony’s recommendations are bi-directional, taking into account not just the actions and preferences of a single user, but also how these actions and preferences align with those of another user and vice versa.
To offer the best customer experience, eHarmony computes and presents matches in real time when a member logs in. The relationship site risks losing customers if there are delays in displaying these matches.
eHarmony’s legacy data store, Voldemort, had trouble meetings the match service’s real-time personalisation demands. Challenges presented by Voldemort included:
- Ever-growing key size. As a plain key-value store, Voldemort could not handle multiple attributes for an object, forcing eHarmony to create keys that were concatenated by every object attribute. These large key sizes made it difficult to maintain the solution and troubleshoot problems that arose
- Less than ideal performance. As eHarmony’s member base grew, so too did the need to serve hundreds of thousands of concurrent users. Voldemort, in no small part due to its key value limitations, could not keep up
- No community support. Originally developed by LinkedIn engineers, Voldemort never received enterprise backing and the project gradually lost momentum. eHarmony felt the data store could no longer be relied upon as a stable enterprise solution
Redis’ footprint expands
Initially used as a user authentication store for eHarmony’s desktop and mobile apps, Prateek Jain, VP and Head of Technology at eHarmony, has bigger plans in mind for Redis. “Redis’ flexible data structures are game changers for our real-time database operations” said Jain. “Rather than generating hundreds of keys concatenated by user ID, time of day and so on, we can use the Hash data structure to greatly simplify the storage of user activity.”
This flexibility has prompted eHarmony to move all its real-time use cases – including key-value lookup table – into Redis and use Redis as its real-time in-memory data fabric. This enables eHarmony to deliver personalised experiences to hundreds of thousands of concurrent customers while they are in session.
Minimising operational overhead with Redis Labs
With Redis playing a larger role in eHarmony’s operations, the company, as part of its ongoing initiative to minimise operational overhead, chose to deploy enterprise-grade Redis, as offered by Redis Labs. Redis Enterprise Pack is downloadable software that provides the high-performance, flexibility and simplicity of open-source Redis, but without any of the attendant complexities that come with maintaining multiple instances of Redis at scale and with high availability.
“Our previous data store, Voldemort, had become too much of an operational burden for the eHarmony team,” said Jain. “The automated management features that come with Redis Enterprise, along with the accompanying 24/7 support from Redis Labs experts, provide the low-touch yet incredibly robust technology platform we were looking for.”
eHarmony and Redis Labs: Future compatibility
As eHarmony gears up to meet the growing needs and expectations of tomorrow’s singles, it considers Redis Enterprise to be a critical component in its solution stack. “We plan to expand our use of Redis Enterprise across eHarmony,” said Jain. “Its ease of use, performance, high availability, flexible data structures and built-in expert support make it highly desirable to us for many additional use cases such as primary data store and message broker.”
