由於公司最近的數據庫服務器CPU非常不穩定。
於是乎下手查找問題的來源。想了下，只能從存儲過程的執行狀態中下手。
查了下資料，發現MSSQL中的系統表sys.dm_exec_procedure_stats會記錄存儲過程的執行狀態數據。字段含義就不累述了。開始幹活：
1、將數據插入一張新表記錄
select convert(nvarchar(10),getdate(),121) as countdate,d.* into procedure_stats_daily from SYS.procedures S JOIN sys.dm_exec_procedure_stats D ON S.object_id=D.object_id
2、每天作業執行，將數據插入記錄表
insert into procedure_stats_daily select convert(nvarchar(10),getdate(),121),d.* from SYS.procedures S JOIN SYS.dm_exec_procedure_stats D ON S.object_id=D.object_id
3、查看時間@Date的執行情況
CREATE PROC CS_RO_Base_GetDailyProcedureStats @Date datetime='' AS BEGIN DECLARE @Date1 datetime SET @Date1=@Date+1 SELECT a.object_id,c.name as '存儲過程名稱', CASE WHEN a.cached_time>=@Date THEN ISNULL(a.execution_count,0) ELSE ISNULL(a.execution_count-b.execution_count,0)END as '執行次數', CASE WHEN a.cached_time>=@Date THEN ISNULL(a.total_worker_time/1000000,0)ELSE ISNULL((a.total_worker_time-b.total_worker_time)/1000000,0)END as '執行時間/秒', CASE WHEN a.cached_time>=@Date THEN convert(numeric(18,4),(ISNULL(a.total_worker_time,0)))/ISNULL(a.execution_count,0)/1000000 WHEN ISNULL(a.execution_count-b.execution_count,0)=0 THEN 0.00 ELSE convert(numeric(18,4),(a.total_worker_time-b.total_worker_time))/(a.execution_count-b.execution_count)/1000000 end as '平均執行時間/秒' FROM ( select object_id,cached_time,plan_handle,isnull(execution_count,0) as execution_count,isnull(total_worker_time,0)as total_worker_time from procedure_stats_daily where countdate=@Date1 )a FULL JOIN ( select object_id,cached_time,plan_handle,isnull(execution_count,0) as execution_count,isnull(total_worker_time,0)as total_worker_time from procedure_stats_daily where countdate=@Date )b on a.plan_handle=b.plan_handle left join sys.objects c on a.object_id=c.object_id order by CASE WHEN a.cached_time>=@Date THEN ISNULL(a.execution_count,0) ELSE ISNULL(a.execution_count-b.execution_count,0)END desc END
根據以上的數據，我們可以查出最頻繁的存儲過程和效率最低的存儲過程，進而給我們的優化工作提供了方向。