Re: I think that relational DBs are dead. See link to my article inside

Hi Josip,

> (In fact I've started writing an essay about that but didn't get much
> interest, http://www.vrspace.org/docs/zen_of_self-adaptive_code-jitsu.html)

It is very interesting. THX!

> And I use soft references so I don't need all objects in memory but keep
> some MRU cached etc etc.

Exactly!

> However, when working with say 10000000 'neurons', I don't even have
> enough RAM for references.

Hm, I am swapping references too. So it is enough to have 1MB to run 1GB network with ... small performance )))

> FTR simple MRU gives me about 97% cache hit rate.

Yes.

It is very strange for me, what i have in experiment. When i disable MRU and load all network into Windows virtual memory (1GB can be loaded), it is many more times worse then when i limit cache to 1 MB and use my own swapping based on MRU (less than 1% cache hit rate with 1MB and full scan). Windows OS swapping kills the performance completly. It takes hours to execute network when using Windows OS virtual memory. And minutes with 1MB RAM to scan 1GB on harddrive.

WBR,
Dmitry Received on Thu Jul 06 2006 - 18:15:13 CEST