Hardware Sensitive Business models

Businesses like Ebay, Search and  Online Dating are all businesses that are not very good from a technology standpoint.   These companies face huge issues because they involve searching and sorting of data.   The more data you get into your system the time to search it grows expotentially.  

You can visualize this by thinking of an excel spread sheet.  If you have a list of a million products all you wanted to find all the ones in blue,  you would have to search each row to see if one is blue.   As the database gets bigger this requires more and more searching.  
This translates into  having millions of computers in the case of google,   17,000 for Ebay and a 1000 servers for a company like match.com.

There have been many many companies in the online dating space that have tried to make a business out of free over the years  but they have all failed and continue to fail.   People forget that as a site grows into the millions of users, the hardware requirements grow expotentially.  This results in the Cost of generating a pageview to skyrocket.   At the same time the revenue per pageview descreases as people start to use the system for fun and their pageviews per session increase.   

Most of the largest dating sites are spending over 1 million a month on hardware and tech staff to keep it running.  The only reason my site is still in business and works so well is because i’ve found a way to solve the technical side of the problem.

Social networking for the most part doesn’t make use of ranged searches.   A profile lookup by ID or username is like looking up row 1000 in a excel spread sheet.  You don’t have to search all the items in the database to figure out what you are looking for, you just want a specific row.    Friendster on the other hand did a search of friends of friends.   This increased the over all computing power needed to keep the site operational expotentially and killed them off.     Myspace on the other hand stripped out everything that would require ranged searches and it became really easy for them to win.

So if you want to make a business that scales really well,  try and make one that avoids searching and ranged searches.

12 Responses to “Hardware Sensitive Business models”

  1. Mayo Says:

    Markus, so basically put a global search like you did i.e. find match in the 70 mile neighboring area and avoid finding friends by name ??

    How did you balance all that, i will be using cluster MySQL (RAM only with disk backup) along with some streamlined PyRex programming and SSD storage (i know your fan of it – it’s damn expensive, but also damn fast :) ).

    Now that you’r big can you share some sauce with us poor wannabe schumcks ?? I’m sure it wont hurt you, well big guy’s wont be capable of using your tech, small guy’s either and i’m entering adult business! ;)

  2. Mayo Says:

    Any good white papers you could mention?? Don’t know where to Google?? :(

  3. Mayo Says:

    How about Eight queens?

  4. Martin Says:

    Markus, how do you handle multi-dimensional searches then? Do you use some tree indexing or composite bitmap indices, or, for searches you use no indices at all except the result ordering index?
    M

  5. Mayo Says:

    Oh BTW, what is your current DB size – both GB and MM of records, and how are you doing in earning — is it up way stream or as before?

  6. Zoltan’s Business News & Blog » Blog Archive » Number of servers Says:

    [...] Number of servers I read at The Paradigm Shift (http://plentyoffish.wordpress.com/2006/11/20/hardware-sensitive-business-models/) that ebay use about 17,000 servers and macth.com use about 1,000 servers. Quite impressive, I wonder how many server admins are needed to manage that many servers. How are these and what exactly is monitorized? How often to they change HDs, memories, network cards, routers, etc.? How often do they upgrade the software? These are all questions that need to be asked before you plan to make a site that has the chances to become … the next internet boom. [...]

  7. Zoltan Says:

    Markus… it looks like you do not want to share this information (DB size) with us. :)

  8. Markus Says:

    80 gigs

  9. Mayo Says:

    :)

  10. Mayo Says:

    Interesting thing i found on Joel on Software:

    http://moishelettvin.blogspot.com/2006/11/windows-shutdown-crapfest.html

    Man oh man…

  11. Mayo Says:

    Wold like you to comment Markus! :)

  12. Wallop = Friendster 2.0? at MINGER.NET - The Long Tail of Web 2.0 Says:

    [...] Friendster had the user experience down pat until it tried to scale when the much celebrated open source platform failed to deliver pages without interminable waits. According to Markus Frind, Friendster on the other hand did a search of friends of friends. This increased the over all computing power needed to keep the site operational expotentially and killed them off. Myspace on the other hand stripped out everything that would require ranged searches and it became really easy for them to win. [...]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


Follow

Get every new post delivered to your Inbox.

Join 243 other followers

%d bloggers like this: