使用LINQ和DateTime对象计算最大并发“用户登录数”

Question

I am having trouble coming up with a proper way to create a query under the following scenario:

I want to find the maximum amount of "users" logged into a hypothetical system at any given point. If, throughout a given period of time, the most peopled logged in at the same time was 5, my query should return 5.

I am using LINQ and would consider myself a beginner; this is the most advanced query I have attempted.

My users are stored in a Database table described like User_Logged_In_Table(user_id, begin_date_time, end_date_time). User_ID is an integer, and begin_date_time and end_date_time are DateTime C# objects.

My current understanding takes me about this far:

myFunction(begin_time, end_time) //<- These two variables represent the threshold of time I am querying about.
{
    var u = from users in database.User_Logged_In_Tables
            where (users.begin_date_time.CompareTo(begin_time) > 0)
                  && (users.begin_date_time.CompareTo(end_time) < 0)
            select users;

            //The above tells me the users who have been logged in between those two times
            //however it does not tell me anything about their concurrency.
}

I am stuck at this point, unsure of how to continue the query with what I need.

Any hints, or suggestions are much appreciated! Also I am willing to further explain my situation if I have been unclear in any area.

Answer 1

So I think I could share this as an answer now. It seems that this is a standard problem, I found something interesting on http://www.haikulabs.com/Schedu10.htm . Basically, your "reservations" are your log-in sessions and your "rooms" are your HTTP connections.

To find the minimum number of connections to handle your users (so, max concurrent users), you would need to test against a DateTime in a for loop and .AddMinutes or .AddHours depending on the resolution you want.

Now, the above won't be accurate - it's possible that there's a short session right between the values you are testing and you'll miss it. So to make it 100% accurate, you need to test against all DateTimes where a session has either started or ended and not against every round minute or every round hour.

Answer 2

"Any given period of time" is vague, so you'll have to lock it down, like by saying you will look at each minute of time in between the start param and the end param, and you will take the average number of concurrernt logins among those minutes to determine the "overall" average.

So now, you'll initialize a list of ints, then loop thru the param start time, adding a mintute for each loop, until you are at the max. Inside that loop, you'll look for distinct logins that have either started in that minute, or are still continuing in that minute. You will grab the Distinct() user_ids (so that one user logging in twice within 5 seconds still counts as 1 login) and you will add that distinct count to the list of int.

At the end, you'll have a list of int that's as long as the number of minutes in between startTime and endTime, and that list can be called with Average() to see the average number of logins for each given minute.

Answer 3

The above solutions for querying based on discrete time intervals is good, I have used the same method in some of my own work applications.

However, one thing to note is that the number of concurrent users only changes when a user logs in or out. Moreover, the number of concurrent users only increases when the user logs in (Sorry for obvious comment, but trying to make sure I have this clear in my own mind too). So it should be sufficient to create a LINQ query that goes over each user found by your initial search (with the range restriction), and checking how many users are logged in at each "login time", and taking the maximum number from this. Actual LINQ below (The code below is in VB.Net, and using my own field names as that is what I had to test with, my apologies if this is a problem):

Firstly, to get a list of concurrent users at each login point we have:

Dim listOfConcurrent =
    From row As DataRow In userRange
    Select New hlxListItem(Of DateTime)(
        row.Field(Of DateTime)("CiDateStart"),
        (From subRow In userRange
        Where subRow.Field(Of DateTime)("CiDateStart") <=
                row.Field(Of DateTime)("CiDateStart") AndAlso
            subRow.Field(Of DateTime)("CiDateEnd") > row.Field(Of DateTime)("CiDateStart")
        Select subRow).Count)

Alternatively, you can get the "hlxListItem" with the highest number to get max concurrent users:

Dim maxConcurrent =
    (From row As DataRow In userRange
    Order By (From subRow In userRange
        Where subRow.Field(Of DateTime)("CiDateStart") <=
                row.Field(Of DateTime)("CiDateStart") AndAlso
            subRow.Field(Of DateTime)("CiDateEnd") > row.Field(Of DateTime)("CiDateStart")
        Select subRow).Count Descending
    Select New hlxListItem(Of DateTime)(
        row.Field(Of DateTime)("CiDateStart"),
        (From subRow In userRange
        Where subRow.Field(Of DateTime)("CiDateStart") <=
                row.Field(Of DateTime)("CiDateStart") AndAlso
            subRow.Field(Of DateTime)("CiDateEnd") > row.Field(Of DateTime)("CiDateStart")
        Select subRow).Count)).First()

So, finally just some notes on the above. hlxListItem is just a simple object with a generic field, and a text field. The generic in this case holds the date (login time) and the text field holds the number of concurrent users. If I have time, I can try to edit the above examples into C# and to using your field names, but I posted the above because I could confirm that it works. Let me know if the above helps you out, or if you need any additional help.