We have implemented a system which uses 'swappable' in-memory data,
binary trees and hashtables to do things like:
o Return 'best' possible route from one person to another based on
weight of relationships between links
o Return 'shortest' possible route from one person to another
o Return total persons within N degrees of seperation
and so forth.
The in-memory data are flushed when memory is tight or when they get
too old, and get replaced with more data from InnoDB tables.
Essentially, it is quite rate to need to swap or load db data.given
that the data structures are not heavy in memory requirements.
Its writen in C++. Our tests with 10 million users (random set)
indicated less than 1 second response time for any of the above
operations (when no mem-disk swaps take place ).
On Wed, 2 Feb 2005 13:44:52 -0500, Balazs Rauznitz <balazs@stripped> wrote:
> On Wed, Feb 02, 2005 at 10:19:45AM -0800, Bruce Douglas wrote:
> > dathan...
> > given that you work at friendster, aren't you kind of restricted from commenting
> on how one would go about creating/implementing this kind of system??
> I thought about this too when describing my company's solution
> (http://multiply.com/ just for full disclosure). Considering that
> technology is just one of many keys to the success of a company, I
> think techies benefit much more from sharing some experiences with
> peers rather than keep everything secret and hope that the other guy
> screws up. Of course as long as help is reciprocal...
> MySQL General Mailing List
> For list archives: http://lists.mysql.com/mysql
> To unsubscribe: http://lists.mysql.com/mysql?unsub=1
Head of R&D
Phaistos Networks, S.A