I'm programming a Web-site and use NHibernate, ASP.NET and MS SQL Server 2000.
Some pages are ready but I'm worring about the performance.

The pages doesn't contain any hard queries.
They just retrieve object and their properties (single properties - links to other object and collections of objects) via NHibernate.

All collections (sets and bags (for sort)) are lazy (have lazy="true")
and not inverse (!!!) (inverse="false")
and biderctional (have link on child side)

In class hierarchies I use string type for disriminator.
(May be it is better to use numeric?)

I created SQL Server 2000 database from the SQL script, generated
by NHibernate (Configuration.GenerateSchemaCreationScript) .

As I correct understand this script contains only indexes for primary and foreign keys and I should create all other indexes (Am I right?).

Tha main question is: what steps should I follow to improve the perfomance of the application.
Some pages execute 10-12 sec.
Sometimes the memory used by aspnet_wp.exe grows up to 100 Mb.


I can:
create indexes for discriminator fields and fields, which
are used in order by and other queries (I can see all SQL queries in NHibernate log)

What do you think is it a good way to start with?
Are other settings in NHibernate to improve them?

Best regards,
Sergey

Zero, profile the application and find the hot spots.
First, get a DBA to look at the databases.
Second, investigate caching.

It's very possible that you're doing things like:

int numOfPosts = blog.Posts.Count;

Which would load all the posts collection to memory just to get its count.

The answer for performance questions is always: Measure & Find Out.

Thank you!
I look at nhibernate log.
At queries but didn't find any bottlenecks.

See a lot of queries with 3-6 left outer join.

---
about "int numOfPosts = blog.Posts.Count;"

I often iterate through collection attributes
(foreach) but never calculate count.

all collections are lazy.

A'm intersted in to know something about standard
bottlenecks of NHibernate configuration.

For example about using of sorted colections
or something else.

Do you know anything about "standart strategy" of creating
indexes in relation DB by xml-mapping files.

Something like: when I see a bas with order-by field A
so I should create an clustered index on A.

I had the same problem and I found that using left join fetch in my queries for the many-to-one or one-to-one relations improved a lot my app's performance. Using caching is also a most.

regards
Simon

Does fetch join mean initializing collection attributes in one query?

what do you mean by caching?
Where can I read something about caching?

Yes



http://wiki.nhibernate.org/display/NH/Enabling+Second+Level+Cache

If you dont know what cache mean

http://en.wikipedia.org/wiki/Cache