[EMAIL PROTECTED] :
> Some years ago I developped a relational database that was faster than
> Filemaker.
> There were about 700000 records... FMP just chocked on it. HC stacks choked
> after
> 12 MBs of data where in...
>
> The trick was to split the interface/engine from the records which were in
> turn split
> into categories/levels, where any category was another stack.
>
> This keeps the searches real fast. The smaller the file, the better...
>
> Secondly, I knew already where the categories to be searched were so I had a
> table
> of doubly linked references between records. This allowed to browse the
> database at
> any point in seconds (Filemaker too over 15 minutes PER search)... This is
> basic
> computer science... Search for binary trees, etc... doubly-linked records...
>
> It's all in how you design your database... Access times were all under 2
> seconds
> even from the cd media!
>
> The reverse of the medal was that importing the records and linking the
> records took
> over 3 days of work non-stop on a Mac 8500 with HC 241. But it doesn't mean
> that
> a real time program couldn't do it real fast. MC being much faster will also
> help.
Yes, this is exactly the kind of trick I used in my recent experiments :
splitting the data into numerous text files, multiple indexing, etc...
The only problem is when you offer this kind of solution to a client :
if you mention Access and ASP (for instance), most ppl in the internet /
CD-ROM biz will take you seriously.
But if you start the kind of above description, you'll just look like a
weird hacker and will probably loose the client...
JB
Archives: http://www.mail-archive.com/[email protected]/
Info: http://www.xworlds.com/metacard/mailinglist.htm
Please send bug reports to <[EMAIL PROTECTED]>, not this list.
