from Hacker News

ArangoML Pipeline: Common Metadata Layer for Machine Learning

by noah-kun on 9/25/19, 1:16 PM with 5 comments

• by noah-kun on 9/25/19, 4:10 PM

  Arango is doing so much right now--their SaaS version "Oasis" coming out this year, SmartJoins for faster multi-cluster joins and now ML.

  I admit I liked Mongo for it's document storage. But I didn't like Postgres due to my irrational dislike of SQL Then I challenged myself to try out a graph database.

  ArangoDB presents things like Mongo (as documents), has a much nicer query language than SQL (or just use GaphQL) and does graph stuff; all while being amongst the top 3 or so in benchmarks that are designed to test only specific types of databases. If databases did triathlons, ArangoDB would be unbeatable.

  Only two issues that I struggled with are/were Kubernetes cluster stuff, like sharding, but this is common to cluster DBs I think.

• by joerg84 on 9/25/19, 2:10 PM

  One of the authors here: Feel free to checkout the meetup talk https://www.meetup.com/Knowledge-Graphs-Meetup/events/264962... and the corresponding slides: https://docs.google.com/presentation/d/1r4rLhjoDN_CWpjMUz29s...

• by janemanos on 9/25/19, 1:39 PM

  Here is the repo on Github https://github.com/arangoml/arangopipe