Question

我们有一个查询，在缓存之前为我们网站上的每个用户运行。对于newrelic，查询占所有问题的大约16％。

这不是一个非常好的查询，我们应该更改表格设计，但我们需要快速修复，以便我们有时间重写此功能。

查询有时需要3秒才能运行。

这是查询。

SELECT 
   q.QuickLinkID
   , q.QuickLinkUID
   , q.SchoolID
   , q.Url
   , q.Name
   , q.Target
   , q.OnlyTeachers
   , q.UserNumber 
FROM 
   Quicklinks AS q WITH (NOLOCK) 
LEFT JOIN 
   schoolParticipants AS sp WITH (NOLOCK) ON q.schoolID = sp.schoolID 
WHERE 
   (sp.userID = ? OR q.UserNumber = ?) 
   AND (q.schoolID = ? OR q.SchoolID NOT IN (SELECT qe.SchoolID 
                                             FROM Quicklinkexclude AS qe WITH (NOLOCK) 
                                             WHERE qe.userID = ?)) 
   AND ((sp.isTeacher > ? AND q.OnlyTeachers = ?) OR q.OnlyTeachers = ?) 
ORDER BY 
   q.Name

Answer 1

将子查询更改为连接并在ID列上应用索引：

SELECT q.QuickLinkID
     , q.QuickLinkUID
     , q.SchoolID
     , q.Url
     , q.Name
     , q.Target
     , q.OnlyTeachers
     , q.UserNumber 
FROM Quicklinks AS q WITH (NOLOCK) 
LEFT JOIN schoolParticipants AS sp WITH (NOLOCK) ON q.schoolID = sp.schoolID 
LEFT OUTER JOIN QuickLinkexclude AS qe ON q.SchoolID = qe.SchoolID
WHERE (sp.userID = ? OR q.UserNumber = ?)
    AND qe.SchoolID IS NULL
    AND qe.userID = ?
AND   ((sp.isTeacher > ? AND q.OnlyTeachers = ?) OR q.OnlyTeachers = ?)

子查询可能会导致表扫描。使用连接，如果您使用正确的索引，它可以是搜索，在大多数情况下速度要快得多。

未经测试，但这应该有效。

t-sql查询占用了大约16％的servertime

1 个答案: