http://www.infoq.com/presentations/Why-I-Chose-MongoDB-for-Guardian

They compare concepts between traditional relational databases and MongoDB.

Although specific to MongoDB, most NOSQL solutions have equivalent concepts, including RavenDb.

Have you switched all or part of your data needs to a NOSQL solution? I’d like to hear how those implementations went, what issues you faced and how much of your data is still relational and why?

• S – Start. (OnPreInit)
• I – Initialize. (OnInit, InitComplete, OnPreLoad)
• L – Load. (OnLoad)
• V – Validation. (Validate)
• E – Event Handling. (OnLoadComplete, OnPreRender, OnSaveStateComplete)
• R – Render. (Render)

Oh and … U – Unload. (OnUnload) but of course SILVER sounds nicer.

This table may look something like this

CREATE TABLE OrderedItems

(

Item VARCHAR (50),
Sequence INT

)

Now, this initially is populated in sequence like so:

INSERT  INTO OrderedItems
        ( Item, Sequence )
VALUES  ( 'A', '1' ),
        ( 'B', '2' ),
        ( 'C', '3' ),
        ( 'E', '4' ),
        ( 'F', '5' )

You can see I deliberately left a gap for item ‘D’.
If I was to insert ‘D’ as a new item and made the sequence 6, what I really want now is to re-sequence the table, ordering by the Item.  If the column used for holding the sequence is part of a constraint or key of the table this isn’t easy. So how can we achieve this?
Well, as it turns out, we can use common table expressions to very easily do this.

WITH LFUpdate AS (
select *, ROW_NUMBER() OVER (ORDER BY Item) AS NewSeq
from OrderedItems

)
UPDATE LFUpdate SET Sequence = NewSeq ;

What do you think, useful?

For accuracy by far the best datatype to store geograhic information is geography, if loaded correctly you can get very neat distance calculations between points, shapes etc based on the curvature of the earth. Nice.

However,  performance-wise – its a dog. Even with spatial indexing applied; over thousands of loaded shapes and having to use STIntersects, STDistance to get the information we need it really wasnt performing.

So, we converted all the shapes, points to geometry – much like this:

DECLARE @geo geography -- This is our geography, imagine its got a shape in

-- Convert it to a geometry

SELECT GEOMETRY::STGeomFromWKB(@geo.STAsBinary(),4326)

All is good in the land of performance, sacrificing slight accuracy (in the UK, this is not significant). We now gain some nifty geometry only methods like STWithin, STContains plus many more. We can do all sort of funky stuff like show me all the postcodes in (STWithin) London. Neato.

Then of course we wanted to apply a radius around our searching, e.g. show me all the Towns within a 3000 metre distance from a postcode… and you cant, at least not as you think.

If the data was still a geography, you’d just use STBuffer(@radius) to apply a X metre radius around your starting point. Now the data is a geometry, STBuffer still exists but its now on a different scale – you cant just (easily) convert your metres to a value that STBuffer will give you an accurate result.

So, can it be done? Well, I think I found an easy answer. By flipping the shape from a geometry to a geography and back we can apply a STBuffer in metres around our shape.

This can be done like this:

DECLARE @radius INT -- e.g. 3000 metres
DECLARE @geo geometry -- my starting shape
SET @geo = GEOMETRY::STGeomFromWKB(GEOGRAPHY::STGeomFromWKB(
@geo.STAsBinary(),4326).STBuffer(@radius).STAsBinary(),4326)
SELECT Geo FROM GeometryTable
WHERE Geo.STWithin (@geo) = 1

 What this little snippet would do is return you all shapes from your geometry table where they are within (inside) your starting shape plus a 3000 metre radius. The calculation is simple and quick. Assuming you have got a spatial index (in our case, we have a bounding box of the UK) the performance is very nippy indeed. Next up; showing distance in metres...

Edit: After this article was published , we ended up tranforming the data via a different project (Tranverse Mercator) which meant we didnt need to flip the geometry to get a buffer.

I have been grabbing the source and compiling each iteration of RavenDB since early on in its development and have seen unbelievable performance and feature changes since. I truly think Ayende has created a substancial contribution to the ever growing market for non-relational databases. I very much doubt Facebook or Twitter will be moving over to RavenDB due to platform/technology choices but its definitely one to watch out for or try for yourself.

Dont know anything about non-relational databases, stuck in MSSQL, Oracle, MySQL land? Well, check out this very good resource on NOSQL and the various database types that fall under this banner.

http://nosql-database.org/

The main four being:

• Document Store (This is what RavenDB is)
• Key Value / Tuple Store
• Graph Databases
• Object Databases

If you want to come along and see RavenDB live, the event details are as follows:

http://skillsmatter.com/podcast/open-source-dot-net/ayende-rahien-introduction-into-raven-db

Curious about RavenDB?; go get, compile the source for RavenDB at http://github.com/ravendb/ravendb

Alternatively, read the comprehensive docs at http://groups.google.com/group/ravendb

It works fine for standard datatypes; float, int, varchar etc. This may also be an issue for other binary based data but I have only tested this with spatial (geography) data.

However, there comes a time when maybe you just want to select some data and do that as easily as possible – most small to medium sized sites probably require basic querying and really dont need some whizzy dba to get the last drop of performance out of the db.

I have been very impressed lately by Subsonic, in particular Subsonic 3 as it introduces the best hands-off approach to databases out of all the ORM tools out there. In version 3 of this mighty tool , it comes with something called the SimpleRepository . What this does is really take away the need to worry about the database. You simply create a class to represent your object (table) and when you select, insert with that object it creates (migrations) the database objects if they are not there. It also handles changes, performaing an ALTER on the fly.

For more information on this , please visit http://subsonicproject.com

http://www.evolus.vn/Pencil/

http://www.simple-talk.com/sql/sql-tools/reporting-services-with-style/