With the rise in popularity of Google Search Appliance and Elasticsearch, many companies are interested in the best search solution for them. At Spantree, we've been asked several times if the GSA or Elasticsearch is a better solution, and we have decided to cover the various strengths and weaknesses. While both are solid search tools which can meet most needs, they specialize in very different domains. To understand the strengths and weaknesses, it's important to note what the general philosophy of each technology is.
Elasticsearch was built upon Apache Lucene, a large open-source project. Both are centered on document storage and retrieval, focusing on searching and sorting many different fields within a document. For example, searching for a list of books would allow you to search and sort by author, title, keywords, etc. In Elasticsearch, this specification can be explicit, by specifying the author, or implicit, by searching for the author's name, results will be returned that contain the desired author.
Elasticsearch was designed to solve the scalability issues many who used Apache Lucene were facing. Shay Banon was hoping to extend his existing library (Compass) to be more scalable, when he realized it required a new solution entirely. Elasticsearch was built from the ground up to be a scalable search, providing the power of Apache Lucene across multiple cores.
One of the risks of latching on to an open source project, especially one with pretty much a single developer, is the danger of abandonment. Luckily, in 2012 a company was created around Elasticsearch, providing more financial and developer support, and adding to the ever-growing features list.
Google Search Appliance
GSA was Google's extension of their own product into the business marketplace. Many businesses desired the same power and consistency that Google offers, but for their own sites and files. GSA has very much so captured the feel of a private Google, as the business that purchases it acquires the actual hardware box, which is used as the search server. This represents the most tangible difference between Elasticsearch and GSA
Because of its roots in typical Google Search, GSA focuses on full-text search. While Google immediately brings to mind standard static web pages, it can also successfully crawl files and databases on the network. This feature, combined with the same security and user management Google offers with their other products, many companies index both private and public records.
Google has continued to upgrade it over the years, both by directly improving the hardware and via the upgrades of their search algorithms. While it may not be their flashiest product, it's consistently given support, features steady improvements, and shows no signs of slowing.
Structured vs Unstructured Data
The two products primarily differ in the way they represent the data. The GSA works with web pages and other documents which are essentially composed of flat text. On the other hand, Elasticsearch documents are JSON objects with a nested hierarchy, which are useful for indexing data that has a lot of inherent structure.
As an example, you can think of two types of data sources. One is a database of employees at a company and the other is a site full of employee biographies. If the employee biographies are just plain text, it may be hard to determine where in the text to find an employee's name or when they started working at the company. On the other hand, this information could be readily available in a database.
Since the GSA is a little bit of Google in your infrastructure, it is unbeatable at finding a whole document that is most relevant for a certain query. But, since it treats documents as objects, Elasticsearch shines at finding some subset of a document that is most relevant for a query. Of course, this oversimplifies things, as both solutions can be massaged into handling both kinds of data. But it is the ease with which you can deal with one sort of data or the other that is important to keep in mind when making a decision.
While the GSA can meet many needs, it comes at a price. Google licenses the boxes at $15,000 / year, with a minimum of a two year contract. Elasticsearch, on the other hand, is completely free and open source software (supported by a multimillion dollar company).
This fact alone will rule out the GSA for most people. If simply interested in a very basic web search, I'd recommend checking out Google Custom Search which is a hosted search solution, but far less robust than GSA or Elasticsearch.
Overall, Elasticsearch and GSA can both solve most problems, but to make sure the best tool is used, evaluate the type of data that needs to be indexed. If most data follows a clear structure, then Elasticsearch is the way to go. If most data follows the form of a webpage, with a few keywords but primarily freeform text, then GSA should be considered.
Kevin is a Senior Software Engineer, and the primary math nerd. After brief stints in physics and academic math, he realized the creation of things was just as important as the theory, and became a full time programmer. Possessing a hodge podge of tools for any job, he particularly enjoys Clojure, Elixir, and pretty much every data store in existence.If you ever want to math out, Kevin would love to talk about cryptography, algebra, graph theory, or number theory. In his spare time, his preferences and hobbies tend to change with the moon, but generally board games, RPGs, and some form of creation lie at the heart of them all.