How does Google make searching so fast?

Behind the scenes of Google Instant

Google Instant

This article is brought to you in association with LG Optimus 2X

Since it launched last August, Google Instant has been a revelation. Its ability to instantly search depending on what you're typing has changed the way many people deal with search results when using a PC or Mac.

After all, you can refine search results with the tap of a few keys rather than having to do a second search when your first isn't so refined. As you'll read, Google Instant predicts what you might want to search for depending on what you're typing rather than what you've already typed.

Its future isn't just restricted to the Mac or PC though. Google Instant has come to Android and iOS devices as a beta preview - you can check this out by visiting the Google search page on your device and turning on the beta.

Google instant beta

BETA OPTION: You can turn on Google Instant beta by going to the Google homepage in your iOS or Android browser

But Google Instant isn't actually that new - in concept at least. It has a long history - going back eight years. Teams at Google have been experimenting with mocks and prototypes of a search feature that returns results while the user types since at least 2003.

"These prototypes faced a number of design and technical challenges," explained a Google spokesperson to TechRadar. "More recently a working research group was formed to investigate these ideas and explore alternatives. That group came up with a version of the idea that closely resembles what we eventually launched."

"As with most launches, we experiment with different features to try and figure out the optimal user experience. Some things work, others don't, but we're very pleased with the Instant experience we've got today."

"Imagining the future of search, the idea of being able to search for partial queries or provide some interactive feedback while searching has come up more than a few times," said product manager John Effrat. "Along the way, we've even built quite a few demos."

Rethinking Instant

Indeed, while Nikhil Bhatla put together a demo in 2003, Amit Patel built something similar back in 1999, not long after Google's launch.

"Our search-as-you-type demos were thought-provoking - fun, fast and interactive - but fundamentally flawed," explains Marissa Mayer, vice president for Google Search Products & User Experience.

"Why? Because you don't really want search-as-you-type (no one wants search results for [bike h] in the process of searching for [bike helmets]). You really want search-before-you-type—that is, you want results for the most likely search given what you have already typed."

Google instant in action

IN ACTION: Google Instant in action on an iOS device

So how does it actually work? As you type, a list of queries are displayed in a drop-down menu below the search box so you can get to the search result you need as quickly as possible. Google dynamically renders a list of relevant search results for the first auto-complete in the drop down menu and as the user continues to type we will show a new results page for new auto-complete predictions.

It uses Ajax - a technology that enables a portion of a web page to be updated without having to update or redraw all the rest of the page. Google performs an Ajax GET request with every keypress.

What's even better is that Instant predicts the rest of your query and shows it in light grey before you finish typing, so if what you need appears you can stop typing and find what you're looking for.

You can also use the arrow keys to move between search results and hit return to go to the link you want, speeding up the basic process of searching.

"New results pages will appear as the top predicted search terms update," continued our spokesperson. "This can happen as the user types a single letter, or a few letters. We've written algorithms to predict when showing a results page is most likely to be relevant to the user."

"Technically demanding to launch at scale"

"Even though we are serving more search results pages, the additional load this enhancement creates is very small when compared to other types of popular web services such as streaming video and online gaming.

"We've also worked hard to minimise the amount of data that is sent and received during the search process. For example, when rendering new search results as people type, we only send the parts of the page that change without updating the static elements, such as a the page frame around the results."

Google admits Instant "was a technically demanding feature to launch at scale". The search giant estimates that, based on an individual search, it is serving between five and seven times as many results as before.

As Google was already processing more than a billion searches per day, this presented a significant challenge. Fortunately it found a number of clever ways to address the greater overhead.

Google deployed new Server Caches that can handle high request rates while keeping results fresh as it continuously crawls and re-indexes the web. It also built systems to adaptively control the rate at which we show results pages in proportion to how relevant the pages are likely to be.

There was also some optimisation of page-rendering JavaScript code to help ensure web browsers could keep up with the rest of the system.

Other tactics include caching information from users. "We introduced user-state data into our back-ends to keep track of the results pages already shown to a given user - this way we don't re-fetch the same results repeatedly."

Google Instant is designed so that it doesn't cause too much overhead on slow internet connections, but it does automatically turn off on very slow connections. And, should you not want it, you're always able to turn Google Instant off should you wish.

And that's the story of how Google Instant - first dreamed of 12 years ago - made its way onto your desktop, your tablet and your mobile.

Article continues below