Granted, if you already run a hadoop cluster you should be able to leverage it for better scaling MapR jobs. Maybe some day Tableau will allow us to write data-source access plug-ins. You'd have to build custom ETL to load data from ES to another DB and query that directly (or query the initial source, if possible). So, yeah, there is some serious overhead and translation / abstraction involved (obviously) and queries will be much slower than native - but the only other (direct) alternative is. So: Tableau -> Hive SQL -> Hadoop -> Map/Reduce -> ES While primarily created to get Hadoop data INTO ES (assumably) we also use it to create an external "table" (more like 'data structure template') in Hive pointing to an ES index and SQL our little goat hearts out on it (and use a pretty generic Hive driver in Tableau to connect to it.). But therein lies a problem - if I'm storing a shit-ton of data in ES, I certainly can't use my go-to visual analytics tool Tableau on it, since ES querying is strictly RESTful. At the very least why would you bother to using something like MongoDB (or a similar "hot" NoSQL sink) when ES is all that and more (plus a joy to work with and scale out). My current love affair (romantic fling?) with ES not-withstanding - it generally isn't usually used as a general purpose data-store (even though it's more than capable), but with the impending release of v1.0 I can see this changing and it's role expanding. (yes, I said "Hadoop" - don't be afraid, little Ricky) Generally speaking, that is (can you use "generally speaking" after dropping an f-bomb?). But in all honestly, to me, it's really a very scalable, auto-balancing, redundant, NoSQL data-store with all the benefits of a full search and analytics server. Put (very) simply, it's a badass search server based on Apache Lucene. I've been a big fan of ElasticSearch the since last Spring - using it on my RiffBank project as well as various other "data collection" experiments.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |