Entity framework and linq to entity performance

Question

I've been investigating using Microsoft's entity framework for a project at work. I created a simple little data layer solution which has an IDataService interface so I could write a standard ADO.Net implementation and an Entity Framework version using Linq-to-Entity.

I've created two tests which request exactly the same data, but use the different implementations. The queries are simple, they retrieve data from a table, and using hierarchical information generate a DTO with the data in a hierarchy.

The data in the database is along the lines of

------------------------
ID  | Description
----|-------------------
1   | Item 1
2   | Item 2
3   | Item 3
4   | Item 4
5   | Item 5

----------------
Parent | Child  
-------|--------
1      | 2
1      | 3
3      | 4
1      | 5

Desired Output
--------------
Item 1
|-Item 2
|-Item 3
| |-Item 4 
|-Item 5

And so the queries currently take the form of:

from a in tableA
join b in tableB on b.Parent equals a.ID
where b.Parent == root.ID
select new DTO.Entry {
  Id = a.ID
  ...
}

The method containing this query is run recursively until there are no more child elements left to process.

Using Linq-to-entity the test takes about 320ms to complete, using ADO.Net the tests takes about 8ms!

Is this just something I have to live with/consider or should the performance be about on par? Also, as the underlying data structure has no referential integrity (I know!), so I am compensating for this in my ADO.Net stuff, but I can't with Entities, is this likely to have an impact?

At the moment it seems that if you want performance then you should stick with ADO.Net

Can you add more information, as in what was the query and what is the datastore? — Lynn Langit
– Lynn Langit, Commented Aug 19, 2012 at 19:39
and what version of entity framework are you using? BTW an interesting link considering performance measurement: blogs.msdn.com/b/adonet/archive/2008/02/04/… — MUG4N
– MUG4N, Commented Aug 19, 2012 at 19:47
It's EF 4.5 (I'm using VS2012 so I'm guessing that's right?). Thanks for the link, I'm scanning over it now — daz-fuller
– daz-fuller, Commented Aug 19, 2012 at 19:53
@daz-fuller: the EF creates objects out of the query results. do you by chance do the same in the ado.net test? also, do you perform like 10 tests and take the average value (to compensate the warm-up of the EF)? — Wiktor Zychla
– Wiktor Zychla, Commented Aug 19, 2012 at 20:01
The ADO.Net instance is executing the query, iterating over the reader and creating the object from each record. I've not accounted for warm up time, I'll re-run the tests and see how that affects it. — daz-fuller
– daz-fuller, Commented Aug 19, 2012 at 20:03

Gert Arnold · Accepted Answer · 2012-08-19 21:48:06Z

Ask a customer: do you want more performance or less bugs?

Of course plain ADO.Net performs better than a data layer that also uses ADO.Net but before and after that does a lot more. Moreover, you chose an area where EF is not a good competitor when it comes to efficiency: recursive queries. But generally, when it comes to writing correct, stable code that performs acceptably, EF (or any seasoned OR mapper) rises head and shoulders above ADO.Net. It is hand-written SQL and data driven programming vs. linq and object oriented programming.

Nevertheless, you right when you say

At the moment it seems that if you want performance then you should stick with ADO.Net

Sure, for performance-critical operations OR mappers tend to have too much internal overhead to fit the bill. And to return to recursive queries: nothing beats recursive queries with CTE's in the database.

Collectives™ on Stack Overflow

Entity framework and linq to entity performance

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related