There is a phrase I use all the time: “Search is one of the most misunderstood functionalities in IT”. And I think it is very accurate.

The problem lies in two different aspects:

1. Developers don’t know how to use search engines. And it is ok, search engines can be hard to tune appropriately and it is a specialised niche. In some cases, there are some search engines which are awfully expensive.
2. Developers are lazy. Let me explain this one.

Let’s say that I am setting up an application for selling cars. Potential customers always look for the same things, which are make, model, year, sort by price and so on and so forth. There is a set of meta data that is important and required to find what you are looking for. So what is the solution to this problem?

Use a database where each field is stored in a separate column and look for the fields accordingly, just like in the following image. It is a mistake or at least a UX horror. I hate database driven search, but that is just my personal opinion.

The correct wat of doing it is by providing a single search box. How? Like this:

If you want to learn how, please click on the following link to my Pluralsight course to get started with enterprise search using Apache Solr!

pluralsight.com/training/courses/TableOfContents?courseName=enterprise-search-using-apache-solr

I have been a fan of Microsoft technologies all my life, probably because I’ve spent a lot of time working with .Net and related technologies. Eventually I became also an Apple fanboy as some people have called me.

But something that I haven’t been called a fan of is Linux. Don’t get me wrong, I think Linux is extremely important, but in my case I have not worked with it as much as I think I should have.

But now I am in a part of my life where I need to run Apache Solr in a production environment. What do I do? What comes naturally.

In a nutshell I set up a Windows machine in Amazon AWS, install Java, download Solr, java -jar start.jar, modify solrconfig.xml, modify schema.xml, turn around a few more knobs and test. Once I am happy I install Tomcat and voila, I have a single node for production. It is a small application with very few documents and a reasonable traffic, so it is all good. And besides, it is amazing how much a Solr instance in AWS can handle.

Anyway, my need keeps growing and I believe I need to set up a more resilient installation. Of course SolrCloud comes to mind, but I am thinking of how the pros install Solr.

So what do I do? Install Solr in a Linux AMI. Also, as I need monitoring now in place I set up SemaText. One downside of Windows is that at least when using SemaText, you can’t monitor on Windows, only Linux.

And there you go, that is my piece of advise. But not only from me, I’ve heard from many sources that Linux can be more performant and stable when running Apache Solr.

If you want to get more information on how to install Solr in a Linux instance, please follow the following link to the Apache Solr Reference Guide: https://cwiki.apache.org/confluence/display/solr/Taking+Solr+to+Production

Also, if you want to learn more about Getting Started with Enterprise Search with Apache Solr, please follow this link to my course on this subject:
www.pluralsight.com/courses/discussion/enterprise-search-using-apache-solr