3 ms·

My source for the costs of Google: http://www.datacenterknowledge.com/archives/2014/07/23/from-112-servers-to-5b-spent-on-google-data-centers-per-quarter/ http:

My source for the costs of Google: http://www.datacenterknowledge.com/archives/2014/07/23/from-112-servers-to-5b-spent-on-google-data-centers-per-quarter/ http://www.datacenterknowledge.com/archives/2014/07/23/from-... "Google data centers cost the company more than $5 billion in the second quarter of 2014" My source for the costs of Bing: http://www.businessinsider.com/microsoft-bing-losing-billions-2011-4 http://www.businessinsider.com/microsoft-bing-losing-billion... "Put differently, Bing is spending $5.5 billion a year to generate $3 billion of revenue." (I divided by four to get quarterly) GPS data is not really on the same level as Web Crawl data. How often is a street added or changed versus a web page?

abalone 12y ago
1. $5B is super way off. First of all you are lumping one-time construction costs in with the incremental costs of running the data centers. The article clearly states the bucket in which data center operational costs are included is $2.82B. 2. Even that bucket is a grab bag of other costs as well such as amortization of acquired assets (not necessarily for data centers). Additionally while web search is major it is certainly not the only thing that Google data centers do; for example they run Gmail and an AWS competitor on it. So it's very hard to determine just what the operational cost of web search is, but it's certainly well under $2.82B. 3. The Bing article clearly states its major cost increases are "costs associated with the Yahoo! search agreement and increased traffic acquisition costs", i.e. promotional not operational costs. This is not applicable to Apple, who already gets paid for traffic. 4. As for maps it's not just GPS data, it's road and signage and building data including visual imagery. That's many orders of magnitude more expensive to gather than loading a web page (even loading it 1000 times). You have to roll trucks and fly planes. In fact there was just a story yesterday about Apple rolling trucks to start gathering competing "streetview" data.. all from scratch.
- bobajeff 12y ago
  Oh, my mistake. You may be right about Maps data being more expensive, if so then it does make sense for apple to make a Web search engine.
- bobajeff 12y ago
  Chatting about this with you made me challenge my knowledge of how web search works. Many articles I've found suggest that the hardest problem to solve isn't competing on the size of Google but the relevance of Google. Which is partly to do with the size and power of their datacenters. But also to do with smart engineering of the servers, database, and algorithms. The algorithms not only determine the how to order the results but which pages to crawl and how often. What this all makes me think of is all the work being put into AI research lately. IBM, Google, Facebook, Baidu etc. There is a race to AI and the winner could very well make a better Google.

abalone 12y ago

1. $5B is super way off. First of all you are lumping one-time construction costs in with the incremental costs of running the data centers. The article clearly states the bucket in which data center operational costs are included is $2.82B. 2. Even that bucket is a grab bag of other costs as well such as amortization of acquired assets (not necessarily for data centers). Additionally while web search is major it is certainly not the only thing that Google data centers do; for example they run Gmail and an AWS competitor on it. So it's very hard to determine just what the operational cost of web search is, but it's certainly well under $2.82B. 3. The Bing article clearly states its major cost increases are "costs associated with the Yahoo! search agreement and increased traffic acquisition costs", i.e. promotional not operational costs. This is not applicable to Apple, who already gets paid for traffic. 4. As for maps it's not just GPS data, it's road and signage and building data including visual imagery. That's many orders of magnitude more expensive to gather than loading a web page (even loading it 1000 times). You have to roll trucks and fly planes. In fact there was just a story yesterday about Apple rolling trucks to start gathering competing "streetview" data.. all from scratch.

bobajeff 12y ago
Oh, my mistake. You may be right about Maps data being more expensive, if so then it does make sense for apple to make a Web search engine.
bobajeff 12y ago
Chatting about this with you made me challenge my knowledge of how web search works. Many articles I've found suggest that the hardest problem to solve isn't competing on the size of Google but the relevance of Google. Which is partly to do with the size and power of their datacenters. But also to do with smart engineering of the servers, database, and algorithms. The algorithms not only determine the how to order the results but which pages to crawl and how often. What this all makes me think of is all the work being put into AI research lately. IBM, Google, Facebook, Baidu etc. There is a race to AI and the winner could very well make a better Google.