Mam tabelę SQL Server 2008 R2, której struktura schematu wygląda następująco:
CREATE TABLE [dbo].[CDSIM_BE]
(
[ID] [bigint] NOT NULL,
[EquipmentID] [varchar](50) NOT NULL,
[SerialNumber] [varchar](50) NULL,
[PyrID] [varchar](50) NULL,
[MeasMode] [varchar](50) NULL,
[ReadTime] [datetime] NOT NULL,
[SubID] [varchar](15) NULL,
[ProbePosition] [float] NULL,
[DataPoint] [int] NULL,
CONSTRAINT [PK_CDSIM_BE]
PRIMARY KEY CLUSTERED ([ID] ASC, [EquipmentID] ASC, [ReadTime] ASC)
WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF,
IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON,
ALLOW_PAGE_LOCKS = ON) ON [MonthlyArchiveScheme9]([ReadTime])
) ON [MonthlyArchiveScheme9]([ReadTime])
CREATE NONCLUSTERED INDEX [idx_CDSIM_BE__SubID_ProbePosition]
ON [dbo].[CDSIM_BE] ([SubID] ASC, [ProbePosition] ASC)
INCLUDE ([EquipmentID], [ReadTime], [BECorr])
WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF,
SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF,
ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [MonthlyArchiveScheme9]([ReadTime])
CREATE NONCLUSTERED INDEX [IX_CDSIM_BE_ProbePosition]
ON [dbo].[CDSIM_BE] ([ProbePosition] ASC)
INCLUDE ([SerialNumber], [SubID])
WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF,
SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF,
ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [MonthlyArchiveScheme9]([ReadTime])
CREATE NONCLUSTERED INDEX [IX_CSDIM_Readtime]
ON [dbo].[CDSIM_BE]([ReadTime] ASC)
INCLUDE ([EquipmentID])
WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF,
SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF,
ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [MonthlyArchiveScheme9]([ReadTime])
I wykonuję to proste zapytanie:
Select Max(Id)
From dbo.CDSIM_BE
Tabela zawiera ~ 2,5B wierszy.
Plan zapytań pokazuje skanowanie indeksu wykonywane na IX_CdSIM_BE_ProbePosition
indeksie. Zastanawiam się, dlaczego SQL Server po prostu nie używałby indeksu klastrowanego (i podstawowego) i od razu przejdzie do ostatniego wiersza w tabeli i pobierze wartość Id, ponieważ musi to być wartość maksymalna.
ReadTime
więc nie można użyć PK, jak opisano. Musi znaleźć Max(Id)
dla każdej partycji, a następnie znaleźć ich maksimum. Możliwe jest przepisanie zapytania, aby uzyskać plan, o którym tu mowa,
select top 1 Id from dbo.CDSIM_BE order by Id descending;