SQL索引表联接

Question

I am executing the following two queries against a SQL database from within my C# MVC application.

Query1

SELECT tableone.id, name, time, type, grade, product, element, value
FROM dbo.tableone INNER JOIN dbo.tabletwo ON dbo.tableone.id = dbo.tabletwo.id
Where name = '" + Name + "' Order By tableone.id Asc, element

Query2

Select DISTINCT element FROM dbo.tableone
INNER JOIN dbo.tabletwo ON dbo.tableone.id = dbo.tabletwo.id
Where name = '" + Name + "'"

Upon running the method that executes these queries each query hangs and oftentimes the next page of my application will not load for over a minute or it will time out on one or the other. When I run the same queries in SQL Server each of them take between 10 and 15 seconds to run which is still too long.

How can I speed them up? Ive never created a SQL index and Im not sure how to create it for each of these or if thats the right path to pursue.

Tableone currently has 20808805 rows and 3 columns, tabletwo has 597707 rows and 6 columns.

Tableone

id(int, not null)
element(char(9), not null)
value(real, null)

Tabletwo

id(int, not null)
name(char(7), null)
time(datetime, null)
type(char(5), null)
grade(char(4), null)
product(char(14), null)

Answer 1

Firstly, as @ Robert Co said, a index on tabletwo.name will help on performance.

Also, are there indexes on tableone.id and tabletwo.id? I will asume there are, given they look like primary keys. If not, you definitely need to put indexes on them. I can see tableone to tabletwo is a many-to-one relation, which means you probably don't have a primary key on table one. You seriously need to add a primary key on tableone, such as tableoneid, and make it a clustered index!

I think another reason here is, your tableone is much bigger than tabletwo which is limited down even further by the where clause(name = 'Name'). This means you are joining a large table (tableone) to a small table (tabletwo with the where clause). In SQL, join large table to a small table is going to be slow.

The solution that I can think about is, maybe you can move some columns, such as 'type', to tableone, so that you can limit tableone into a small set in your query as well :

Select DISTINCT element FROM dbo.tableone
INNER JOIN dbo.tabletwo ON dbo.tableone.id = dbo.tabletwo.id
Where tableone.type = 'some type' and name = '" + Name + "'"

I am not quite sure how these suggestions fitted into your data model, I just hope they may help.

Answer 2

10 to 15 seconds with 20 million rows and no index? That's not bad!

As Ethen Li says it's all about indexes. In an ideal world you would create indexes on all columns that feature in a filter (JOINs and WHEREs) or ORDER BYs. However, as this could severely impact UPDATEs and INSERTs you need to be more practical and less ideal. With the information you have provided I would suggest creating the following indexes: CREATE INDEX index1 ON tableone (name); If tableone.id is your candidate key (that which uniquely identifies the row) you should also create an index on it - possibly clustered, it depends how ID is generated):

CREATE UNIQUE INDEX IX1TableOne ON tableone (id);

Or

CREATE UNIQUE CLUSTERED INDEX IX1TableOne ON tableone (id);

For tabletwo: the same applies to ID as for tableone - create at least a unqiue index on ID.

With these indexes in-place you should find a significant performance improvement.

Alternatively to add primary key constraints:

ALTER TABLE tableone ADD CONSTRAINT pktableone PRIMARY KEY CLUSTERED (id);

ALTER TABLE tabletwo ADD CONSTRAINT pktabletwo PRIMARY KEY CLUSTERED (id);

On tableone this might take a while because the data might have to be physically re-ordered. Therefore, do it during a maintenance period when there are no active users.