我有一个脚本,该脚本必须查看超过250万条记录,以查找是否有未读电子邮件的成员。我想知道可以采取什么措施来提高其速度。目前,运行脚本最多可能需要8秒钟的时间:
SELECT TOP(1) MemberMailID FROM MemberMail WHERE ToReadFlag = 0 AND ToMemberID = 102 AND ToDeletedFlag = 0 AND FromDeletedFlag = 0 AND OnHold = 0 AND ToArchivedFlag = 0
如何使用索引使其更快?
该索引可能会有所帮助,但是请记住,这里没有免费的午餐(必须维护索引,因此这会影响您的插入/更新/删除工作量):
CREATE NONCLUSTERED INDEX unread_emails ON dbo.MemberMail(ToMemberID) INCLUDE (MemberMailID) WHERE ToReadFlag = 0 AND ToDeletedFlag = 0 AND FromDeletedFlag = 0 AND OnHold = 0 AND ToArchivedFlag = 0;
现在,您的查询可以说:
SELECT TOP (1) MemberMailID FROM dbo.MemberMail -- dbo prefix WITH (INDEX (unread_emails)) -- in case you need to force, though you should not WHERE ToMemberID = 102 AND ToReadFlag = 0 AND ToDeletedFlag = 0 AND FromDeletedFlag = 0 AND OnHold = 0 AND ToArchivedFlag = 0 ORDER BY ToMemberID; -- ORDER BY is important!
如果您根据查询更改其中一些标志的值,则可以尝试将这些列添加到索引键而不是过滤器中,例如,假设有时您要检查OnHold = 0,有时OnHold = 1:
OnHold = 0
OnHold = 1
CREATE NONCLUSTERED INDEX unread_emails ON dbo.MemberMail(ToMemberID, OnHold) INCLUDE (MemberMailID) WHERE ToReadFlag = 0 AND ToDeletedFlag = 0 AND FromDeletedFlag = 0 AND ToArchivedFlag = 0;
您可能还想尝试使用MemberMailID键而不是INCLUDE。例如:
MemberMailID
INCLUDE
CREATE NONCLUSTERED INDEX unread_emails ON dbo.MemberMail(ToMemberID, MemberMailID) WHERE ToReadFlag = 0 AND ToDeletedFlag = 0 AND FromDeletedFlag = 0 AND OnHold = 0 AND ToArchivedFlag = 0;
这些差异可能对您的数据和使用模式无关紧要,但是您将能够比我们所能猜测的更容易地测试差异。