Enthusiastic for Elasticsearch
March 01, 2018
Enthusiastic for Elasticsearch
Score 10 out of 10
Vetted Review
Verified User
Overall Satisfaction with Elasticsearch
We decided to start looking into Elasticsearch after we had good success with using lucene (the full-text search indexer that Elastic uses). We had some queries in Oracle that were running EXTREMELY slow and knew we had to do something for the customer to make their experience better. We had a few thoughts on what we could use and Elasticsearch fit what we really wanted.
- Searching, it does it well and searches are fast...real fast.
- Ease of use, we were able to get an Elasticsearch cluster up and running in a half hour and doing basic searches after that was very easy with simple requests
- Redundancy built in and stability. We haven't had any of our Elastic clusters go down intentionally, but testing out redundancy by removing nodes Elasticsearch has gone flawlessly.
- Only breaking changes between versions when they are absolutely necessary.
- Works well with .Net libraries that are supported and coded by Elastic.
- A bit more of a learning curve for complex searches, indexing more complex things.
- Some of our updates between versions haven't gone as smoothly as we would like, but in more recent versions Elastic has done a much better job at trying to allow for full uptime upgrades.
- Configuration needs to be set up to do larger searches, or more complex searches and at times while starting it wasn't obvious what configuration needed to be changed.
- The first and highest reason for switching to Elasticsearch was to speed up the queries that we had that were running slowly(full text search over millions of records). We had some Oracle searches that were taking upwards to 45 seconds. After switching to Elasticsearch those same exact queries were running under half a second. It was obvious to us what the return on the investment was there.
- The first thing(unexpected but made sense) that we noticed when switching to elasticsearch was our database servers didn't need as many cores. As you pay for Oracle licensing by cores, this was a huge benefit. We dropped about 6 cores in our Oracle licensing as soon as we could after switching to ES.
- Since we had such great success with searching one table we decided to include more tables into our searching to help with our database.
The only other competitor we researched was mongo as some of our table information is stored in an XML file, but as we were doing searching we gravitated towards Elasticsearch. We knew mongo had some of the qualifications for what we wanted, but went with Elasticsearch for specifically our searches and actually used Mongo for more DB storage.