How to count a very large volume of request

[Arsen V.]

Hello,

What is the best way to accomplish the following.

1) An ASP.NET program (consiting of one file somepage.aspx) receives about
25,000,000 requests pay day.

2) The requests come from a limited set of IP addresses.

3) How to count the total number of request from each IP address per day.

SQL Server 2000 is used on the backend.

Currently we used the following architecture:

* Each request to somepage.aspx generates an INSERT into a LogTable1
* There is a clustered index on the LogTable1 on the dateStamp field which
is of the type smalldatetime
* The dateStamp field has a default that sets it to the getdate()
* At the end of each day at 12:01AM there is a simple query that runs and
does a group by to count the number of requests from each IP in the given
date range (past 24 hours). This query works great and takes only 2 minutes
to run.

Is there a better way to accomplish this without having to do INSERTS into
the LogTable1 for each request?

It would not work to have the ASP.NET program execute an UPDATE each time to
increment the total number of request, since this would cause LOTSSS of
locking in the database layer.

Thanks in advance.

Arsen

 

[Kevin Spencer]

You could store the data in your Application Cache, and update it
periodically to the database.

--
HTH,
Kevin Spencer
..Net Developer
Microsoft MVP
Big things are made up
of lots of little things.

 

[Ignacio Machin \( .NET/ C# MVP \)]

Hi,

What about using the IIS log file?
you could set it in the way you need it, basically the client IP and maybe
the time, if you configure it to update daily all you have to do at 12:01
is run a process that read the file generated and do what you need.

Cheers,

 

[Arsen V.]

Hi Kevin,

Do you suggest storing the TOTALs in the Cache?

Or storing the actual requests: date and ip

Would there be a locking problem?

How to do the "periodic updates" to the database from the Cache?

Thanks,
Arsen

 

[Kevin Spencer]

Hi Arsen,

You could put a DataTable in the Application Cache, and add records to it
with each Request. As for periodic updating, you could put a routine in the
Session_OnStart Sub that checks an Application DateTime variable, and at
certain intervals, inserts all the records from the DataTable into the
database and clears out the DataTable.

To be safe, you would want to add code to your Application_OnEnd sub to
update the database if the Application stops or times out; however, with 25M
requests per day, that might not be necessary.

--
HTH,
Kevin Spencer
..Net Developer
Microsoft MVP
Big things are made up
of lots of little things.

 

[bruce barker]

what i have done is cache the last 15 minutes. the cache has the ipaddress,
the start of the 15 min interval, pagename, and number of hits during the
interval. as most users cluster their hits, this cuts down the number of
inserts.

i flush the cache every 15 minutes of when too large.

you can then get daily or hourly stats from the db with simple queries.

-- bruce (sqlwork.com)