Transaction Isolation Level In SQL Server: A Complete Guidance

Daniel Jones | January 19th, 2017 | SQL, SQL Transaction Log


Transaction isolation level on SQL server, decides or determines that how transaction integrity is apparent to other systems and users. The locks (acquired to protect data modifications) are not affected by the transaction level, which is chosen. An exclusive lock is assigned for the transaction of any data, which holds or modifies the lock data without depending upon the isolation level, which is set for the transaction. For read operations, isolation level on SQL server primarily exemplify the level of protection made by other transactions.

A lower transaction isolation level increases the number of concurrency effects like, dirty reads or lost updates, which can be encountered by the users. However, it also increases the ability (of many users) to access the data at the same time. Whereas, a higher isolation level lessens the types of concurrency effects, which a user may face. A higher isolation level increases chances of the blocking of one transaction to the other and requires more system resources. The highest isolation level fetches the same data without any modification every time it repeats the read operation. While, the lowest level can read uncommitted data. It can also, gives back the data that have been modifies, which is not committed by other transactions.

There are five transaction isolation level in SQL Server

  • Read Uncommitted
  • Read committed
  • Repeatable Read
  • Serializable
  • Snapshot

Understanding Different Transaction Isolation Levels In SQL Server

Read Uncommitted

A query in the current transaction can read the data modified within another transaction, which is not yet committed. This is a list restrictive isolation level because it the database engine does not issue the shared locks when Read Uncommitted is specified. It is possible to modify the data by any other transaction between issuing statements within this transaction, which results in non-repeatable or phantom reads.

Read Committed

This isolation level prevents dirty reads as a query in the current transaction level cannot read the data, which is modified by another transaction that is not committed. However, non-repeatable or phantom reads are still possible as data can be modified by other transactions between issuing statements within the current transaction.

Repeatable Read

A query in the current also, prevent dirty reads. The reason behind this is that it cannot read data that is modified by another not committed transaction. Moreover, this transaction does not allow any other transaction to modify the data until it completes. Therefore, avoids non-repeatable reads.


A query in this transaction cannot read modified data by any other transaction, which is not yet committed. Also, the data cannot be modified by any other transaction read by this current transaction. In addition, any other transaction cannot be inserted new rows until the transaction completes. Therefore, Serializable isolation prevents phantom, dirty and non-repeatable reads.


A statement can operate the data only when it will be in a consistent state during the transaction. Once the process started and any other transaction modifies the data then, the data is not visible to the current i.e., Snapshot transaction. This transaction does not block other transactions from writing data, neither do this transaction request locks while reading the data.

Note: It is not possible as the default isolation level for all SQL server database is Read Committed. So, if you want to use another transaction isolation level then only option is to set it manually within a session.


Management of data is very necessary and for this SQL proves to be a major platform. Transaction isolation level on SQL server makes this task very easy and convenient. In the above discussion, we have learned that there are five transaction level on SQL with their proper understanding.

Transaction Log Restore