Data Tier Optimization

[Mervin Williams]

I am building a web application that will have a single form that will be
populated with data from 8-10 tables. The easiest way to implement the data
tier would be to use DataSets and Data Adapters for each table. But, is
that the optimal approach?

How do I optimize the back-end data retrieval and corresponding updates
(inserts, updates, and deletes) within the .NET Framework?

Thanks in advance,

Mervin Williams

 

[William Ryan eMVP]

Mervin:

If it's all on one form (and even if it's not) having a seperate dataset for
each datattable is definitely something you don't want to do. One DataSet
holding the datatables, (Unless logically there's a good reason for using
seperate datasets - but if you're talking about one form, it's hard to
imagine a scenario where you'd want/need more than one dataset), using
datarelations where possible is probably the best approach. You can then
serialize one set if you need to , pass it to/from a web service or use
whatever other method you want to access the data. You'll want different
dataadapters.

As far as optimization, there are different schools of thought, but you can
encapsulate that in any tier although the presentation tier is usually a
really bad choice. If you use stored procs you cna keep it all in the
backend layer which I find preferable b/c I can make change to my program
without recompiling and redistributing and it allows for optimum
performance, secuirty and flexibility. Many disagree and think the logic
should be in the middle tier and they have a point. You can also combine
the two. However, calling procs, passing in params where needed and letting
them do the heavy lifting is a way that I prefer.

HTH,

Bill

 

[Guest]

I am curious when is it evident to break up a dataset with mulitple tables.
For example if you have multiple forms and need data on each form. Do you
suggest to break up the one dataset logically to match the forms if the
forms map to schema of the db? Or one dataset is still better? I have seen
many examples using only one dataset and it seems to me that this is not a
good approach.

Thanks