At the forefront of big studies and you will machine training, eHarmony’s matching system uses a rhythm formula running on a giant selection of hosts to help you process good billion potential fits each day international. Built on MongoDB, eHarmony’s “possible meets creator” uses bi-directional, user-discussed standards to complement users according to an elaborate, total number of attributes and you may choices.
MongoDB’s versatile data model and you may strong indexing have permit eHarmony to help you manage multiple-characteristic looks so you can quickly narrow down what number of prospective suits to own single men and women. The procedure of identifying fits are narrowed out-of 15 days to have the whole member pool in order to twelve days having MongoDB, a great 95 % reduced big date.
“With these prior, decentralized SQL-dependent program, the whole account put try stored on each servers, and therefore affected efficiency and you may impeded all of our capability to level horizontally,” said Thod Nguyen, CTO at the eHarmony.