Is elasticSearch still useful when using only one-node (server)?

Question

I'm curious but Couldn't find any resource regarding a one server (node) elasticsearch system. I have read about scaling-out benefits of NOSQL and it data-sharding. if running my NOSQL db on a one server and having elasticSearch on same server, would elastic still be faster than the alternatives (relational db with normal queries/ lucene search) or would it will be 'loosing' it advantage as a distributed/clustered search engine?

As I'm not much interested in the NOSQL-search VS. SQL-search but more into the replication/distributed style of Elastic and it Disadvantages of using it in a one-node-cluster.

Thanks

Answer 1

You won't lose performance by running it in a one node cluster - in fact it may be faster. You do lose redundancy, reliability and ease of scaling out if needed. But since it no longer has to sync the master's meta data with other nodes in the cluster it would likely be faster than if you had two nodes... Also ES actually sits on top of Lucene. Of course with one box you lose your replication since there is nothing to replicate with!

That being said are you looking for NOSQL or document storage? ES was originally intended as a document store with built in fuzzy search. I know a lot of folks (including us) are using it as a distributed NOSQL-type DB but there are faster platforms for doing this if you don't need the extra bells and whistles ES comes with.

Finally, you should consider how the JVM (and specifically GC) plays with the other services running on this box.

If this is a potential production system I'd be concerned. Don't run big services on the same box, even if you have to split them up to two smaller boxes.