Google Instant Search and energy savings

Google Instant Search ?

Since mid September, Google has added a new feature in its search engine : “Google Instant Search” ; this feature updates the search results in Google while you are typing your query.
google instant search snapshot

I am not going to discuss if this feature is convenient or not (I personally don’t think so because of the delay between the query typing and the results display I often don’t even have time to click it – I’d rather type slowly :-) ), I am only wondering what is the impact of this function on the overall power consumption in Google data centers ? My first impression is that we will multiply by 5 the amount of energy required to perform a search when using Google Instant Search since every time we enter a new character the result section is refreshed.

Let’s talk about energy figures

Early last year most of us discovered, thanks to the Harvard University physicist Alex Wissner-Gross, how impressive is the energy consumption of Google with a very simple figure : 2 researches on Google generate almost the same amount of CO2 (carbon dioxide) as boiling a kettle for a cup of tea !! Since Google handles about 10 billions requests per month we have 5 billions kettles per month !!!
Google, on its official blog, amended this figure to say that every query only represents 1kJ (0.0003 kW) of energy. Then one month of search on Google would require 3 000 MW (roughly 2 nuclear power plants…), hopefully distributed on the entire month.

What has changed with Instant Search ?

How does it work ?

Instant Search uses AJAX (Asynchronous Javascript And XML) so that the client can get results without reloading the whole page. To keep it simple, when you reach Google page your browser download a piece of a program that enables asynchronous communications between your computer and the remote server. Try to deactivate Javascript in your browser to see the difference !
So even if you never spotted it before you send request to Google server even if you don’t press the “Search” button as long as you type some text for your query.
More precisely, the normal behaviour can be clustered into 3 steps:
1) after you have written the first character, Google anticipates what will be your request, it sends a request and displays some answers (you can see the proposal in light grey). The proposal is calculated by Google algorithm probably considering your history and most popular requests.
google instant search snapshot

2) then you can continue typing, as long as you stick to what Google predicted your request will be initialy.

3) but every time we deviate from the prediction, Google recalculates what you request might be based on the current sequence, then a new request is sent to the server and the results section is refreshed. From the intial 9 millions results we now have only 1 million.

So basically we can say that we sent 2 requests instead of only 1 !


Let’s assume we are looking for a plumber :
– “P” –> request for Pandora
– “L” –> request for Playlist
– “U” –> request for Plugged-In
– “M” –> request for Plum
– “B” –> request for Plumbing
– “E” –> request for Plumbers
( at the time I was writing this)
I don’t want to consider errors in typing but they should be !
So instead of 1 single request for “Plumber” we had 6 ! I am not saying this will multiply by 6 the consumption of the data centers since it is probably non linear but for sure it won’t be neutral.

Where is the famous Google Green policy then ?

Apart of the numerous arguments highlighting According to Google communication, this feature should reduce the time we spend to perform search by 2 to 5 sec. So the assumption is then we will spend 350 millions hours less searching on Google.
If we convert this amount into energy (with a 600W standard equipment) we get a saving of 350 000 000 * 600W = 210 000 MWh / year.
If we multiply by 3 the amount of energy which is required for a search because we use Instant Search we have an overconsumption of 3 000 MW * 12 * 3 > 1 000 000 MW / year.
So at the end it doesn’t seem balanced on the energy side.
On the economic side it is even worst, when you know that Google already represents 2% of data centers consumption in the world and spends $38 millions in electricity per year you might think why they are likely to consume more energy for us to consume less :-s

By the way it is quite funny that Google announced at the same time his awesome its offshore wind power project that should produce 6 000 MW.

All that stuff seems a bit odd to me but I am sure they did more complex calculation to ensure they will get something, they always succeeded in such things.
I am also confident they will continue in promoting green technologies : solar panels, hybrid cars, data centers energy savings, … because they always been in advance on this topic.

Knowing more in details how it works exactly and what are the benefits for Google would be interesting but I guess it is currently a hot topic to discuss since Yahoo argues that this technology is licensed to them.

About Matthieu

Hello, I joined Java-Hoster in 2009. After my engineering studies in France, I worked for several companies including some well known Natixis , Airbus and Air France . Now I am more focused and involved on Java open-source projects and of course in Java-Hoster which makes me learn a lot and feed me as well :-D I am also responsible for this blog, so if you have something to say about it feel free to contact me. Cheers.
This entry was posted in search engines and tagged , , , . Bookmark the permalink.

2 Responses to Google Instant Search and energy savings

  1. Anonymous says:

    I hope you will keep updating your content constantly as you have one dedicated reader here.

  2. click here says:

    Although I really like this publish, I think there was an punctuational error near to the finish from the 3rd sentence.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>