Getting timeout expired. Timeout period elapsed prior to obtaining connection from connection pool with Dapper - dapper

Question

One application is giving exception timeout period elapsed prior to obtaining connection from connection pool. I have used below code, this code snippet is called from different concurrent users and can have maximum hits up to 10000 pr.second. I have used Dapper for fetching the results from Azure MS SQL database.
public async Task<List<Results>> GetDBResults(string Id, int skip, int take)
{
var values = new DynamicParameters();
values.Add("Id", Id);
values.Add("Skip", skip);
values.Add("Take", take);
using var connection = GetConnection(AppSettingsProvider.TryGet("ConnectionString"));
try
{
var sw = Stopwatch.StartNew();
//connection.Open();
// QueryAsync is from Dapper
var dbResult = await connection.QueryAsync<ResponseObject>("SP_name", param: values,
commandType: CommandType.StoredProcedure, commandTimeout: Constants.CommandTimeout);
var result= dbResult?.ToList();
Console.WriteLine("execution time = {0} seconds\n", sw.Elapsed.TotalSeconds);
return result;
}
finally
{
connection.Close();
}
}
private SqlConnection GetConnection(string connectionString)
{
var sqlConnection = new SqlConnection(connectionString);
return sqlConnection;
}
I know, 'using' will close and dispose the connection object, Connection is getting closed but DB pool is not available immediately to next DB connection. So, I have closed DB connection explicitly in final block. This made me to execute few more concurrent request successfully. After that, I am getting timeout exception.
Connection.Open is managed by Dapper, so no connection.Open is added to the code snippet.
We are getting timeout issue after concurrent users crossing more than 200 hits.
Please let me know the resolution to this problem.

Answer 1

This would appear to be a pool exhaustion issue; if you have 10000 requests per second, and noting that the default connection pool size is 100, then even if everything else works perfectly (never CPU starvation, GC stalls, etc), then you only have 10ms for your "acquire, execute, release" loop (since each of the 100 connections needs to handle 100 requests per second to achieve a total of 10k/s requests). That's pretty ambitious, to be honest, even on hardware with a good LAN - if you're in the cloud ("Azure MS SQL"): forget it (unless you're paying for very very high tier services). You could try increasing the pool size, but I honestly wonder whether some kind of redesign is the way to look here. The fact that you're measuring the time in seconds emphasizes the scale of the problem here:
Console.WriteLine("execution time = {0} seconds\n", sw.Elapsed.TotalSeconds);