Microsoft parts Azure cloud, reveals NoSQL doc database Microsoft has slipped out DocumentDB for Windows Azure, the company’s first-ever non-relational database – and its first new database product since SQL Server. DocumentDB is a complete departure from Microsoft’s relational roots, being a schema-free, NoSQL offering built entirely for consumption as a service on its cloud. …
I think I'm missing something but I don't understand how any NoSQL system can perform a "complex query" and still maintain it's supposed breathtaking performance.
The moment you get away from filtering based only on the key, the system now has to perform a full table scan of all the data in order to find what it wants. At which point, the benefits of NoSQL vs a proper relational system completely evaporate. Unless of course, you introduce indexing... But that can't be right ,because that's an evil thing from the world of relational databases!
Can somebody clarify?
Nothing evil about indexing.
For example, CouchDB lets you write a function in Javascript which pulls out or combines any fields you want from a document, and emit them into an index to create a "view". (Postgres expression indexes are similar).
MongoDB does indexing too. http://docs.mongodb.org/manual/core/indexes-introduction/
Postgres a much better option, available on aws rds or Heruku, if you require fully managed service.
For those too lazy or not intelligent enough - to design properly normalised data structures - pg has simpleton friendly, key value store (hstore) plus json datatype - now binary and faster than mongdb.
You can also store xml and index using powerful functional indexes on xpath functions.
And of course you can still use the far superior Relational Model, for sensible data structures, which are far more flexible and easier to query than old fashioned ( think 1950's IMS) hierarchical structure such as json or xml.
Anyone who thinks it's a good idea - to pre-bake the navigation of information and to keep repeating wastefully their information in a data store, must have not really thought about this or they must be retarded.
Fact is most developers don't understand relational model and can't get hierarchical sequential file oriented thinking out of their tiny brains - Joe Celko describes them beautifully as ID Idiots.
It's so easy to model a document or a graph in sql - ( they are also relations) - so why on earth would any one wish to be so constrained.
A Postgres indexed closure table is as fast as neo4j.
Developers shout the mantra DRY - yet they think jsonb makes sense as a datastore ??? even csv is better - tables of course by far the best - especially now that io speed problem solved by ssd.
Keen to find out what's it like I head over to the Microsoft Document DB site (named like there are no others, just like MS SQL) to download myself an installer.
Turns out, its a Document DB as a Service offering. Say what? Has this world gone stark raving mad? I can no longer develop locally but have to rely on an external party. When will this stop?
argh.
Well, perhaps the first non-relational database since DOS 3, if you count all their other non-relational databases as bought-in from outside. But then SQL Server was bought in from outside too.
[It was DOS 3 that added record locking and data sharing to the file system? ]
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
