At Reversim Summit 2016, AppsFlyer, a SaaS mobile marketing analytics and attribution platform based out of Israel, presented their SingleStore use case. Yulia Trakhtenberg, Data Team leader at AppsFlyer, shared her team’s story and spoke about why the SingleStore in-memory scalable database provided the best solution for their problems.
AppsFlyer aggregates some 8 billion daily events in real time and their solution could not handle this amount of data using their current solution. Their team dealt with issues surrounding unresponsive dashboards, faulty modeling, and difficulty introducing new features. In addition, AppsFlyer faced dangerous failures and the risk of losing mission critical data with no hope of recovery.
AppsFlyer required a database that scaled in real time, comfortably processed more data, and by extension, possessed more data dimensions. They wanted a very stable database that could expand without fear of unanticipated crashing. Even with all of these capabilities, performance still had to be very fast.
In their search for an in-memory database that offers all of the above and more, AppsFlyer tried various options and none provided the desired solution.
They tried Druid, MongoDB, Redis, Cassandra, Amazon Redshift, and more with no success.
Finally, Twingo, a senior partner and an official representative of SingleStore in Israel, introduced the SingleStore in-memory scalable database to AppsFlyer. The new AppsFlyer solution that implemented real-time aggregation over SingleStore integrated with batch processing over Apache Spark.
SingleStore database is recoverable in case of data corruption or loss, possesses the capability to return to 0 point in case need arises to go back and start afresh. Furthermore, SingleStore supports 30 times more data than the previous solution AppsFlyer utilized.
In the end, SingleStore saved AppsFlyer $20,000 monthly compared to the fee they paid to run their previous solution per month. Now, AppsFlyer can store more data, which translates to more money.
The new SingleStore architecture not only addressed the pains AppsFlyer experienced, but allowed them to aggregate 10 times the amount of data with much faster response rates, keep up with product demands, all while remaining a cheaper solution.
Check out the full presentation deck from Reversim Summit: