Object Oriented Database in Azure? | The Coffeehouse

magicalclick

I am thinking of trying Azure, but, I don't know if there is any OO Database offering by MS. Does MS database support OO db? And how does that work? Does it do Relational-OO hybrid? And how does the data stored? Do I still have deplicated foreign keys because of relational DB? I am just getting start to learn OO DB, so please don't expect me to know some basic stuff.

Thank you.

blowdart
Peek-a-boo

No MS doesn't offer one.

It depends what you mean by it of course. You can layer an ORM on top of SQL, linq2sql or the entity framework, or others, and you're getting objects generated for you.

magicalclick

In reply to blowdart

#Oct 5th @ 2:58 PM

Thanks

Does Linq2sql/entity framework offered by MS or I have to get it from 3rd party? And how does the performance match up? And is it worth it? Or I should just stick with SQL?

As far as my impression, OO DB seems to be faster and lower bandwidth. But, somehow I fear there is some fishy draw backs. Like the raw data is not as easy to pull out as table. How does MS do their skydrive and hotmail? I am pretty sure SQL won't be enough.

rhm

In reply to magicalclick

#Oct 5th @ 3:09 PM

Object oriented databases are a discredited concept. I don't know where you're getting this idea that OO DBMs are something that Microsoft would even consider having as a product line. The only one I can think of that is useable is Caché and that's not really what people were talking about as an "Object-oriented Database" in the 1980/90s.

blowdart
Peek-a-boo

In reply to magicalclick

#Oct 5th @ 3:09 PM

Hmm, ok, what are you aiming for here? Persistance of objects? Or just blobs of simply indexed data like google does with bigtable?

staceyw
Before C# there was darkness...

In reply to magicalclick

#Oct 5th @ 3:09 PM

"As far as my impression, OO DB seems to be faster and lower bandwidth. But, somehow I fear there is some fishy draw backs. Like the raw data is not as easy to pull out as table."

Don't think OODB is any faster in general. I would definately take a look at Linq2Entities and/or Linq2Sql. In some ways, gives you best of both worlds.

magicalclick

In reply to staceyw

#Oct 5th @ 6:06 PM

Thanks guys, I will look into those stuff. Join tables sometimes is redundent. There is once I have to do recursive join tables for sales tracking on a particular big customer in a month. Because the order tracking can span all the way to 2 or three months, I didn't limit on how old the tracking should do. I bombed my companies DB untill stop running the report. I later limit the tracking to one month older, but, it is not where need accurate. When recursion happens, the join just gets too big to handle.

Yeah, what's that google bigtable thing? I am actually trying to see how everyone build their DB can scales nicely. My company DB doesn't sacles that well because a lot of time I have to pull both archive and working table, which is a really big action. Specially on those search things. How to design a DB that is good for search? Should I make transactions tables per customer to speed it up? But I don't see how it can be implemented.

littleguru
<3 Seattle

In reply to magicalclick

#Oct 5th @ 7:32 PM

SQL Azure is not a OO database. It is a relational database. But as mentioned, you could put entity framework on top of it. The performance is quite good. There is an overhead introduced by the mapping but it is OK.

spivonious

In reply to magicalclick

#Oct 5th @ 7:32 PM

Why would you have to do any joins for sales tracking?

"SELECT sales_amount FROM sales WHERE customer = ? AND sales_date BETWEEN ? AND ?"

magicalclick

In reply to spivonious

#Oct 6th @ 6:23 AM

It is sales tracking. I am doing,

select * from SOP_fromBothHistoryAndWorkingTable where customerID = "ABC" from date ? and ?

Plus all its related records. Meaning,

if result.PreviousSOPNum != null then get that_record, and if that_record.PreviousSOPNum != null then get thatthat_record, and repeat all the way until thatthatthatthatthatthat.PreviousSOPNum == Null.

And print out all the info from

firstSOP (),

that_record (invoice),

thatthat_record_1 (original order split as backorder), thatthat_record_2 (original order split as backorder),

.........,

thathathathathahtahtahta_record (original order).

Basically I have to list all the Invoices form that customer and plus all the records related to it all the way from initial order. I forgot why they want this report, but, it was crazy. And I am using crystal report, so I don't have the flexibility on SQL statements.

blowdart
Peek-a-boo

In reply to magicalclick

#Oct 5th @ 7:32 PM

So big table, HyperTable, Cassandra, and Azure table isn't a relation database at all. Basically it's a store where you do all the work.

So you have a key column (well two actually), and then, well, and then another column which contains your data. Everything centers around your key selection, which lends itself to partitioning and map reduce really well, but if you get the key wrong then it's never going to be fast.

Eric from UK DPE has a slide deck on Azure storage

magicalclick

In reply to blowdart

#Oct 6th @ 10:55 PM

super thanks for the link, I read it and it is fairly easy to understand. Looks like I really need to sign up for TCP and try it out. So it seem like pulling all the data out from Table first and do it locally is good.