Truncating TEMP table before insert leads to big amount of logical reads on TEMP TABLE

Question

Setup: We have a complex set of queries, that runs quite often (about 3000 times per hour) for different users. At the start of the process, .NET application creates a set of temp tables during the process, .NET application calls a set of SQL queries, they look like this:

truncate #QuerySpecificTable

insert into #QuerySpecificTable
select PrimaryKey
  from Production.QuerySpecificTable
  join ...
 where ...

Issue i came across while performance tuning one specific query. ~139000 records get inserted into the temp table. This causes ~300k logical reads on the #QuerySpecificTable.

When i remove the truncate, it drops to ~300 reads on #QuerySpecificTable. CPU statistics show drop in CPU time (~900->300ms so not sure about precision and reliability.)

The issue is, when trying to do this in SSMS, i cannot 100% reproduce the behaviour and make sure and demonstrate removing the truncate (even through it wouldnt be a problem) is a good thing to do. The biggest issue is NOT KNOWING why and what is happening here

WITH TRUNCATE (NOT ALWAYS, thats the weird part)

truncate table #test
insert into #test select top 150000 PrimKey from dbo.test

Table '#test_______________________________________________________________________ ________________________________________000000000141'. Scan count 0, logical reads 319955, physical reads 0, read-ahead reads 217, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.
Table 'Test'. Scan count 1, logical reads 246, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

(150000 rows affected)

DROP/CREATE

drop table if exists #test
create table #test (PrimKey integer primary key) 
truncate table #test
insert into #test select top 150000 PrimKey from dbo.test

Table 'Test'. Scan count 1, logical reads 246, physical reads 0, read-ahead reads 232, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0.

(150000 rows affected)

Answer 1

The data in there is not really dropped, first of all. It has to do with how Temdb is implemented

Also, Microsoft has a great blog on TempDB and contention that addresses the following points:

Tempdb - Files, Trace flags, updates - Microsoft Blogs

We are continuing to work on improving tempdb performance and metadata contention, but in the meantime, there are some best practices you can employ in your code that might help avoid the contention:

1. Do not explicitly drop temp tables at the end of a stored procedure, they will get cleaned up when the session that created them ends.
2. Do not alter temp tables after they have been created.
3. Do not truncate temp tables
4. Move index creation statements on temp tables to the new inline index creation syntax that was introduced in SQL Server 2014.