This is the diagram:

And this is sql, which, as I understand it, is too complicated for SQL Optimizer:
SELECT * FROM
(
select pp.Id as PaymentPartId, b.Id as BudgetId, grouping(bp.ID) as g1 , sum(pp.Amount) PaymentsSum, sum(bp.Amount) BudgetSum
from Projects pr
inner join Payments p ON pr.Id = p.ProjectID
inner join PaymentParts pp ON p.Id = pp.PaymentId
inner join Budgets b ON pr.Id = b.ProjectID
inner join Budgetparts bp ON b.Id = bp.BudgetId
group by pp.Id, b.Id, rollup(bp.ID)
) x
WHERE x.PaymentPartId = 777
SQLFIDDLE: http://sqlfiddle.com/#!6/aa74e/11 (with auto- generated data)
What I expect: the execution plan should contain an index search on x.PaymentPartId . What for? Since this query is equivalent to:
select pp.Id as PaymentPartId, b.Id as BudgetId, grouping(bp.ID) as g1, sum(pp.Amount) PaymentsSum, sum(bp.Amount) BudgetSum
from Projects pr
inner join Payments p ON pr.Id = p.ProjectID
inner join PaymentParts pp ON p.Id = pp.PaymentId
inner join Budgets b ON pr.Id = b.ProjectID
inner join Budgetparts bp ON b.Id = bp.BudgetId
WHERE pp.Id = 777
group by pp.Id, b.Id, rollup(bp.ID)
... and the last query uses index search.
But SQL Optimizer not only refuses to use the index, but also ignores all the hints (I suggest you expire with sqlfiddle - this is really interesting).
, : , SQL Server Optimizer ? , rollup - , sql WHOLE.
P.S. , ", ": (sqlfiddle !).