Keynote: “Graphs are eating the world” was by Emil Eifrem, who is the co-creator of Neo4J, about what graph databases are and whats expected in Neo4J 2.2+.
He is expecting to have 1 million records per second on the initial import of data, up to 10 times faster cypher queries and up to 100 times faster concurrent writes.
From the sources he has shown, No-SQL databases are becoming more prominent in the enterprise world and its expected that by 2017, 25% of enterprises will be using Neo4J or other No-SQL databases in production.
Oh and the geekiest thing out. The introduction video is Beautiful Day by U2 – using Neo4J where, through the UI, queries are run which in turn, display the lyrics as nodes and relationships. Below is a screen from another song:
“The Business Graph” by Kurt Freytag from Crunchbase, has just released a new global tech events calendar, with a development and QA time of 2 weeks using Neo4j as a backend. He spoke about why Neo4j hasn’t been completely adopted yet and a few reasons by Kurt as well as Volker Pacher related to how the web industry has evolved over the last two decades. Originally, CGI applications were heavily focused on data modelling and having front ends mapping directly to those models. Subsequently, the ORM world came about and that primarily introduced an abstraction layer, which still tied app developers back to data modelling. Graph database development tends to remove this focus on data modelling and allow for easier data manipulation (as opposed to schema changes for relational databases). Another reason is related to the fact that it is all still new and right now, is just isn’t plug and play for web development languages. e.g. MySQL. There is also some overhead in thinking in graphs as opposed to relational normalised databases. There is also a lack of tools at this point in time.
“How eBay Delivers Even Faster Using Neo4j” by Volker Pacher started with him ordering a bottle of scotch on eBay to be delivered immediately. That functionality was done by Shutl which was a company that was bought by eBay that Volker worked for as a developer. Its a delivery/courier company aggregator that collates their data and costs to show the end user a time and cost, and the rest is handled by their API/service. He spoke about moving their architecture to graph databases which subsequently provided them with speeds of 8000ms down to 400ms. Oh, and it did arrive by the end of the presentation.
“Neo4j at Scale using Enterprise Integration Patterns” by Brad Nussbaum was enlightening as he spoke about MediaHound – a company which allows you, the user to download content, be it video, audio, radio, books, to any of your devices. They use a graph database in the backend, but what was most surprising was their use of ESB’s in their architecture. It is primarily used to split up the batches of data that are fed into their system, by first sending them to queues and once the process has finished, then publishing it to a topic for consistency (spot servers were mentioned as, and I’m not sure about this, but sometimes server time is bidded on, and if you’re outbid, then your instance will shutdown and thus you need to make sure that all instances are up to date).
Then there was a panel on using Neo4j in the Enterprise world as well as a closing keynote by one of the developers of Neo4j, Jim Webber, who is a hilarious guy from the UK (I can only guess).
Graph Connect 2014: http://graphconnect.com/