Index principle and filling factor in database

SQL SERVER Index principle and filling factor

stay SQL Server in , The index is by B Trees （ Balance tree ） Structured , Indexes B Each page in the tree is called an inode ,B The top node of the tree is called the root node , The underlying nodes in the index are called leaf nodes , Any index level between root node and leaf node is called intermediate level , When adding new data , To ensure that you need the same number of reads to find each page , Intermediate level pages will be split to generate new layers , As shown in the figure .

The width of each layer is increased to the number of pages that the previous layer can record , When the existing tree cannot record more pages , A new layer will be created , The size of an indexed record is affected by the size of the indexed column , So the narrower the indexed column , The more indexes you can put on a page , Thus, the index requires fewer layers , Each floor needs 1 Sub logical reading , So the fewer levels of the index tree, the better .

The fill factor is the percentage of page fill data for each leaf layer , The fill factor option is provided to optimize index data storage and performance , Use fill factor Option to specify Microsoft SQL Server How full each page is when creating a new index using existing data .fill factor Option is an advanced option , If you use sp_configure System stored procedure to change this setting , Only in show advanced options Set to 1 Can only be changed fill factor, The settings take effect after the server is restarted .

When an index is created or rebuilt , The fill factor value determines the percentage of space on each leaf page to be filled with data , So as to reserve a certain percentage of free space for future expansion of the index . for example , Specifies that the value of the fill factor is 80 Indicates that each leaf level page will have 20% The space for is left empty , To provide space for the extended index as data is added to the underlying table .

The fill factor can be set to 1 To 100 A percentage between , in the majority of cases , The default value for server scope is 0, If the fill factor is set to 0, It means that the entire leaf level page is filled , However, in the actual test process, it is generally not set to fill full leaf pages , Because at least one more index row needs to be added , Use this setting to effectively use leaf level space , However, some space should be reserved for limited expansion before the page has to be split .

Be careful ： Fill because it is set to 0 And set to 100 Same meaning , Means that the entire leaf level page is filled . And only after the index is created or regenerated , To apply the fill factor ,SQL Server Database Engine The specified percentage of free space is not dynamically maintained in the page , If you try to keep extra space on the data page , It will be against the original intention of using fill factor , Because with the input of data , The database engine will have to split pages on each page , To maintain the percentage of free space specified by the fill factor .

If you add a new row to a full index page , The database engine will move about half of the rows to the new page , To make room for this new line , This reorganization is called page splitting . Page splitting can make room for new records , But performing page splitting may take some time , This operation consumes a lot of resources . Besides , It can also cause debris , Which leads to I/O Operation increased , This will directly affect the performance of the database . Choosing the fill factor value correctly provides enough space to expand the index as you add data to the underlying table , Thus reducing the possibility of page splitting . If page splitting occurs frequently , The index can be rebuilt by using new or existing fill factor values , To redistribute data .

Although lower fill factor values are used （ Not 0） Reduce the need to split pages as the index grows , But indexing will require more storage space , And it will reduce the reading performance , Even for applications that target many insert and update operations , The number of database reads generally exceeds the number of database writes 5 To 10 times . therefore , Specifying a fill factor that is different from the default value will reduce the read performance of the database , The decrease ratio is inversely proportional to the fill factor setting . for example , When the fill factor is 50 when , The read performance of the database will be reduced by two times , Read performance is degraded because the index contains more pages , Therefore, the disk needed to retrieve data is increased I/O operation .

What does the fill factor in the index mean

If you have a glass full of water , You will try to add more water to this glass . What will happen ? The water will overflow .

　　SQL Server The same is true of . When the index page is full , If you try to add a new row , be SQL Server About half of the rows will be moved to the new page , To make room for newly added rows . This is what is usually said “ Page segmentation ”. Page splitting makes room for new records , But it is time-consuming and very resource consuming , And will produce fragments , Thus it is possible to I/O Operation has a negative impact . that , How to avoid this kind of situation ?

　　 To prevent this kind of situation , The fill factor value must be determined proactively . After the index is created or rebuilt , The fill factor value determines the percentage of space used to fill data in each leaf page , The rest is reserved for future expansion . for example , Configure the fill factor value to 60 It means that for each leaf level page 40% Are empty , To provide space for index expansion when adding data to the underlying table .

　　 The default fill factor value is always 0, This value applies in most cases . In general , The fill factor value is 0 This means that the leaf level is almost full , But there is some room , At least one more index row can be added .( Please note that , Fill factor is 0 and 100 Is similar to that of .)

sql Index fill factor is the best , What is the use of fill factors ？

The most relevant to index reconstruction is the fill factor . When creating a new index , Or rebuild an existing index , You can specify a fill factor , It is the number of data pages in the index that are filled when the index is created . Fill factor is set to 100 It means every index page 100％ fill ,50％ It means every index page 50％ fill .
   
   If you create a fill factor of 100 Clustered index of （ On a nonmonotonically increasing column ）, That means whenever a record is inserted （ Or modify ） when , Page splits happen , Because there is no space for this data on the existing page . Many page splits will reduce sqlserver Performance of .
   
   for instance ： Suppose you just created a new index with the default fill factor . When sqlserver When it was created , It places the index on adjacent physical pages , Because the data can be read in sequence, this will have the optimal i/o visit . But when the watch follows 、、 Add and change , A page split has occurred . When page splitting occurs ,sqlserver A new page must be allocated somewhere on the disk , These new pages are not contiguous with the original physical pages . therefore , The access is random i/o, Not in order i/o, This makes it slower to access index pages .
   
   So what is the ideal filling factor ？ It depends on the application to sqlserver The ratio of reading to writing a table . The first principle , Follow the instructions below ：
   
   Low change table （ The reading / writing ratio is 100：1）：100％ Filling factor of
   
   High change table （ Write more than read ）：50-70％ Filling factor of
   
   Half read and half write ：80-90％ Filling factor of
   
   You have to experiment before you can find the best fill factor for your application . Don't assume that a low fill factor is better than a high one . A low fill factor reduces page splitting , It also increases sqlserver Number of pages read during query , This reduces performance . Too low a fill factor not only increases i/o expenses , It also affects the cache . When data pages are moved from disk to cache , The whole page （ Including empty space ） All moved to the cache . So the lower the fill factor , Have to move to sqlserver The more pages in the cache , It means less space for other important data pages at the same time , This reduces performance .
   
   If you do not specify a fill factor , The default fill factor is 0, signify 100％ Filling factor of （ The leaf of the index 100％ Filling of , But the middle page of the index has reserved space ）.
   
   As part of monitoring , You have to decide what the fill factor is when you create a new index or rebuild an index . in fact , Except read-only database , In all cases , The default value 0 Are not suitable . contrary , You want a fill factor to keep the right free space , Follow the above discussion .

Reprint ：

SQL SERVER Index principle and filling factor - You know

What does the fill factor in the index mean -CSDN Forum

sql Index fill factor is the best , What is the use of fill factors - Patty green onion - Blog Garden