--- David E Jones <[EMAIL PROTECTED]>
wrote:


> The main trick is caching. You can always measure
> the time it takes  
> to do a query, but in general doing a cache look up
> is generally two  
> orders of magnitude faster than a db query, and as
> you mentioned  
> below the product information doesn't change as
> often so caching  
> works well.
> 

So, aside from a view-entity not actually being
handled by the database by declaring an explicit SQL
"CREATE VIEW...", is it possible to cache the results
of a complex-alias view entity?  If it's cached, is
there then a difference whether it's being handled by
an entity or a view entity on subsequent requests?

> 
> What would the goal be for this?
> 

Just brainstorming, three goals
1) Creating data warehousing techniques where the
proof of concept end result is more applicable to a
wider user base than simply the report generation
that's generally more implementation specific.

2) Decrease the learning curve by separating the
Product entity by the meaning of the data, without
having to scrap much of the work that is in the demo
installation.

3) Normalization.  Decrease the potential of breaking
one feature when introducing another (or creating
roadblocks from new features being created).  

> In general this sort of thing is difficult and
> costly 


I don't think it would be difficult as you'd be using
a normalized data set to create the denormalized
temporary table.  Costly, perhaps.  And that's where
my question is really aimed at. If you had 100,000
products, would it take five, ten, fifteen minutes to
run a service to recreate a new temporary table?  

And even if it did take that long couldn't you avoid
the effect on a live system with the following
strategy.
Product - denormalized product related entity
ProductSub1 - normalized portion of Product Entity
ProductSub2 - normalized portion of Product Entity
...         - more normalied portions of Product
Entity
ProductSubN - normalized portion of Product Entity

Run service that denormalizes portions of ProductSub1,
ProductSub2, ... ProductSubN to create
ProductNext - denormalized product related entity not
in use yet
Drop Product, rename ProductNext to Product

> and makes  
> things like real-time effective dated relationships
> and other such  
> things difficult.
> 

Wouldn't the dating of the relations in the normalized
entities still be correct even though the denormalized
entity is being dropped and recreated?

> -David
> 


Reply via email to