How to choose the number of disk arrays ...
Hello everyone!
Here are some timeless on the separation of database files on disk arrays - ideally controllers - separate ...
We note first that the various configurations suitable for databases systematically isolate the transaction log on a separate set of disks, for two reasons:
- Regarding data security, the total loss of data discs can be recovered using logs transactions, and vice versa;
- Insulating transaction logs provide better overall system performance by reducing queues on different disks.
Then, for reasons of general performance of the database, it is important to set the tempdb on a single disc set ... Indeed, all databases in the DBMS using this base to create the temporary objects during the various queries. This is particularly important in SQL Server 2005!
Also for reasons of performance enhancement, it is desirable to isolate the unordered index (NONCLUSTERED) on a set of separate disk and thus increase the performance of research by these indexes. However, separate CLUSTERED index is totally useless, as it would move all data ...
It is also recommended to move large data - often named for BLOB Binary Large Objects - a set of disks separated to minimize the splitting of data ... For this, the horizontal partitioning of data is needed.
More and more databases have extremely large volumes of data, which can strain the performance greatly. Vertical partitioning of the solution ensures that data is separated on different data spaces ... It is often useful to isolate these games FILEGROUPS on separate disks.
Finally, it is important to remember that:
- Discs systems are not data discs, and must be reserved for use by the operating system and executables,
- disks used for storing backup files where applicable should also be dedicated disks to avoid any degradation of performance of the database,
- And to implement this type of architecture, nothing beats the RAID ... So read on!
0 comments:
Post a Comment