The Web A-Team

The web is getting fast. Technology obsolescence can go very fast in our industry and companies always adapt to meet their clients demand.

This is the case for websites search services.

3 years ago (quite a long time in the web industry), I had to choose a new search technology for a client web site. The site was a custom-built CMS and, as you can imagine, the original search feature were quite basic. The choices we had for a quick integration lead us to Google Mini: it was offering a known interface for our users (a copycat of Google Search with custom tabs) and was very reliable.

In between, I was already developing some sites with Drupal and faced the recurring problems of search with Drupal on medium-big size sites. Then Apache Solr integration module came out and I have since selected it as a 1st choice for internal search. Acquia is offering Acquia Search for the clients not willing to deal with own Solr Server for various reasons and this is a really fair deal if you can't afford your own Solr server.

In this post, I will compare the current offers from Google and Apache Solr integration with Drupal (Acquia Search or dedicated Solr hosting).

1. Google Mini

Pros

• Google Search Experience
• Possibility to use it for Intranet sites
• Indexing of file attachments
• Requires minimum maintenance (updates of firmware mainly)
• Output Results as xml, json...
• Customisation possible through xsl-xml transformation and indexing of metadata information

Cons

• Has to be hosted on your site network (so you may have to pay co-hosting, network traffic etc...)
• This is a black box within your network: you don't have fully access to its interface so even if options are available, difficult to fully take control of its crawling rate (and crawler can actually add some load on your servers)
• Support is not easily available
• Need of xsl knowledge in order to get a good integration with a site
• Difficult to know how this would cope under heavy use (if XML api is used for more than basic search: More Like this features, Ajax calls...)
• Firmware updates are available for 2 years only as part of support, after this, Google kindly suggests you to renew your license and get a new hardware...
• in some contexts, language detection can be tricky (especially on pages which include mixed languages)
• no faceted search out-of-the-box

2. Google Site Search

Google Site Search is less customizable than Google Mini. Its market is the small to medium size sites which want to have a basic search integration. It is some kind of white label Google Search and equivalent to a Google Search restricted to a particular domain with some extra features. 

Pros

• Is a web service, no hardware support required
• Indexes new content fairly quick (as quick as Google does)
• Light customisation which doesn't require specific knowledge

Cons

• Limited integration with specific content
• No faceted search out-of-the box
• Limited to site search, no integration with extended features such as content suggestion

3. Acquia Search - Apache Solr dedicated server

There are 2 options there. Either you do not want to host Solr on your servers and buy Acquia Search service through Acquia Network (you'll have access to their support services for Drupal too) or you install it on a server of your own.

There are plenty of tutorials about how to install Solr on a dedicated server, this one is very good for Ubuntu/ Debian systems: Nick veenhof's blog.

Pros

• Neat integration with Drupal (Solr is notified on key events about node updates or node insert), if cron runs frequently, new content appears within minutes
• Custom design
• Faceted search out-of-the-box
• Various extension possibilities: as usual with Drupal a lot of modules have been developed around Solr (attachment search, local search, multisite search, Ubercart integration, multilingual support)
• Possibilities to develop stand-alone pages based on Solr: Solr doesn't only restrict to searches, listing pages can also be created with Solr, giving fast and responsive user experience

Cons

• If hosted on your server, a fair knowledge of Java application deployment is required

4. Price Comparison (price per year)

I didn't include the potential cost of a standalone installation of Apache Solr on a dedicated server. From my experience, Apache Solr doesn't require too much maintenance if you are familiar with Java application deployment.

If you have a few spare servers in your network, this is a good idea to try it first there. If you are concerned with high availability and support, Acquia Network is a good choice for Solr implementation.

5. Conclusion

Whereas the offers overlap on most points, pricing is not linear for Acquia Network as the assumption is that bigger web sites, indexing a big volume of pages will also benefit from custom support.

Google Site Search will mainly benefit small to medium sites, looking for a quick integration without shiny features and with low integration with their Drupal application.

Google Mini would fit for medium to big sites, looking for an integration with a Drupal site. However, its profile would suit better an intranet, knowledge base integration.

The question is probably where does it fit in between Google Site Search and Google Search Appliance. Apparently Google is relying on the 2 services to take over the niche market of Google Mini.

Apache Solr and Acquia Network are the 1st choice for a Drupal site. They are especially interesting for websites which want to extend basic search features, adding content suggestion etc...

With Drupal 7 and the push for RDFa, the integration of Views for Apache Solr, new possibilities are around the corner. Stay tuned!