Posts Tagged ‘google’

GMail outage and reliability

Tuesday 24 February 2009

Gmail has had an outage today. Being at home and doing some private business, this is a reminder just how accustomed I have become to this service being operational always. Internet connectivity has become a basic utility just like electricity and water. And this happened in a time frame of about 10 years. I got my first ADSL connection back in 1998, obtaining, activation and configuring the line to work was still an experience rather than something which happened daily.

The regular post is a system we have can rely on most of the time, it just works. GMail has the same promise. What we often do not realize is the technology and experience to run these online services is relatively new and experience is not easily found. Google has invented its proprietary architecture to run its search and many of its other services. They make heavily use of three concepts:

  • small units, that are easy to replace
  • redundancy, when one part of the system fails an other can take over
  • compartmentalization, when a failure exists the cause can be easily identified

Incidentally the normal post system uses the same principles to handle the enormous amount of letters each day.

The pioneering work from Google is now becoming a available from commercial vendors and service providers as well. Nick named cloud computing, the complete physical layer from the infrastructure has become an abstraction that well… just works. Always.

Advertisements

Data Center power consumption and CO2 emissions

Tuesday 13 January 2009

One search in Google accounts for 7 grams of CO2,  2 searches is the equivalent of boiling a kettle of water. The accurateness of this claim can be discussed. But if you have ever set foot in a computing data center it is clear that data centers consume a lot of electricity to power the servers and cool the building. As a result data centers have a serious CO2 emission footprint. And the relative high growth of data centers compared to other industries validates a closer look.

As a McKinsey study pointed out:

Already, the world’s 44 million servers consume 0.5 percent of all electricity, with data center emissions now approaching those of countries such as Argentina or the Netherlands. In the United States alone, growth in electricity used by data centers between now and 2010 will be the equivalent of ten new power plants. Without efforts to curb demand, current projections show worldwide carbon emissions from data centers will quadruple by 2020.

And in a graphic some other disturbing facts:

co2datacenters

Some of the steps that are taken to mitigate the growth in electricity consumption:

  1. Consolidating data centers and servers. Virtualization is now becoming ubiquitous even for the most demanding applications. Virtualization is driving down the current inefficient use of servers. 
  2. New facilities that take advantage of new technologies like natural cooling and power supplies that produce fewer.
  3. Raised awareness among regulators to hold organisations accountable for electricity consumptions and emissions operating data centers. 
  4. A global industry consortium called The Green Grid dedicated to developing and promoting energy efficiency for data centers and information service delivery. 

Some companies at least claim to think about the environmental costs of running their data center operations and take action to improve it: Google, IBM

Read more:

Disclaimer: 23 google searches were executed in creating this article.

Search

Monday 6 October 2008

Implementing search in a website has one big problem associated with it:

People think search is like Google and relatively easy to implement. If you have read some of the history about Google, you this is not true: giving good answers to users search queries is very hard. Google has done exceptionally well in hiding something very complex behind a very compelling and easy to use interface. An interface that seems to just work. Google is a perfect example of how users behave, it is not about optimal results it is about good enough results.

So lesson # 1: do not underestimate the complexity of implementing a good search solution for your website. As a consequence think up front how important search is, what kind of search your visitors expect, how you want to use search to guide navigation and how much time, effort, resources and thus money you want to spent.

In subsequent posts I will talk about meta versus free text search, and non organic search algorithms.

A great book about the history of Google is “The Google Story” by David Vise.