Re: Widely distributed searching

[Richard Letts <[e-mail]>]

OSI Directory Service writes....
> 
> >The basis for this belief is that I have added a new feature to the DE
> >interface (available as the PARADISE service) which allows for multiple
> >organisation searching, or power searching as someone has coined it.  This
> >...
> >                                         In the meantime, you can try the 
> >service on one of the PARADISE project's machines.  Do let me know what you 
> >think.
> 
> I think it works great from a user's perspective!  As a DSA manager, however,
> I am concerned about the load if this practice becomes widespread (and why
> would I not want to use a 'power search', now that I know it exists :-)
> 
Agreed; how many of use used the directory to see how many
Barker's/Letts'/Hawthornes's there were in the directory ? :-)

The load it places on the DSA depends upon several factors:
	1. Are the searches optimised
		- looking at the searches produced they are neatly optimised.
		looking for "r letts" produces a search of the form:
		(cn=r**)|(sn=letts) - very neat!

	2. Are attributes optimised, and the subtrees indexed ?
		- they search cn= , and sn= attributes; these are normally
		optimised.

	3. Is your DSA compiled with TURBO_INDEX defined
		- This speed searches by an order of magnitude if (1) and (2)
		is true; which it normally is.

	4. Do you have enough memory in your DSA ?
		- A much longer message will be posted on this. Our
		core-image has grown to 33MB; on a 24MB machine some searches
		take a considerable length of time.

Because of these factors, especially 1-3, the load isn't significant.

Richard
-- 
Richard Letts			My opinions are my own, and do not necessarily
Network Manager                 reflect those of my employer
University of Salford               Email: [e-mail]
United Kingdom			    Phone : [phone]
United States of Europe			<I wish I was paid in ECU>