List of Differences between SQL Server and Oracle

List of Differences between SQL Server and Oracle

1. In SQL Server, only one instance runs per platform.

2. SQL Server tablespaces are called databases, and lie half way between an oracle tablespace and an oracle instance.

3. Five databases always exist: master (=system); tempdb (=temp); distribution (for distributed transactions); model (the template for all new database/tablespaces); and msdb (automatic scheduled jobs information). Three sample databases are also created on install, but can be safely deleted: pubs; northwind; user1.

4. SQL Server databases comprise one or more datafiles, as in oracle, but also have one or more online redo log files. This is the principal difference between SQL Server and oracle; the idea of the temporary tablespace, system tablespace, and each schema tablespace having their own set of online redo logs takes some getting used to.

5. SQL Server has logins and users. Logins are specific for a SQL Server instance. Users are specific to a database. A login will have one user for every database it has access to. Logins have systemic privileges (create database, etc.), Users have DDL and DML privileges. Standard practice is to name users the as their parent login - this is done by default.

6. Users with execute permission on a procedure / select permission on a view will fail to use this if there is broken ownership with referenced objects, unless they are explicitly given access to the referenced table.

7. Revoke (neutral) and Deny (strong) are both available in place of Oracle Revoke. The difference is whether or not they can be overridden by role or user privileges. Deny appears as a red cross, Grant as a green tick, and Revoke does not appear in Enterprise Manager privilege box. [p109]

8. Objects can have their schema changed in SQL Server without being rebuilt, using sp_changeobjectowner object, owner. But SQL has to be rewritten to specify the right user, especially with broken ownership chains, so probably this is more trouble than it is worth.

9. SQL Server does not have synonyms.

10. Standard Practice on SQL Server is to have all objects, even views and stored procedures, owned by the dbo user. Other schemas would tend to have their own database/tablespace. However, this is just standard practice, and does not have to be followed. But problems could arise be with lack of synonyms and broken ownership chains.

11. SQL Syntax is slightly different. SQL Server books online has full syntax with a search capability. However DDL and DBA operations should be done by the Enterprise Manager GUI, which has wizards available. SQL Server Enterprise Manager is vastly superior to the equivalent Oracle GUIs. A few operations cannot be done through Enterprise Manager: Column permissions have to be modified through SQL; Filegroups have to be created through SQL.

12. In DDL, whitespace and special characters must be covered with []. E.g.: REVOKE ALL ON [order details] FROM PUBLIC; DENY CREATE DATABASE TO Eva, [Corporate\ErikB], Ivan.

13. SQL Server has its equivalent of v$ and dba_ views. It also has a collection of system stored procedures which return v$ information or actually perform system DDL.

14. SQL Server has application roles in addition to standard roles. Application roles are password identified while standard roles are always enabled for a user. When an application role is enabled, no other privileges are apparent, with the exception of public privileges. Application roles would typically be enabled and disabled through a VB front end script. Syntax is: exec sp_setapprole 'approle_name', 'password'

15. Databases have a primary datafile (*.mdf), possibly some secondary datafiles (*.ndf); and one or more on line redo log files (*.ldf).

16. There is no reason to give a database more than one datafile, except for backup/recovery streamlining for very large database/tablespaces.

17. SQL Server has a fixed block size of 8k, and a fixed extent size of 64k. This may not mean an end to defragmentation requests, since there is still a reorganise utility. The reorganise utility can be scheduled to run after backup, and does hot reorgs. Small tables can, apparently, share an extent.

18. Rows cannot span blocks, so the maximum row size is 8k. This means chaining does not happen in SQL Server, but problems will occur if a row physically cannot fit into 8k.

19. The default size of on line redo log files is 25% of the total size of all datafiles. It is also recommended that autoextend be switched on on online redo log files.

20. A maintenance wizard will decide whether or not to grow or shrink on line redo log files, among other things.

21. Autoextend can be in extensions of a fixed size or a percentage of current size.

22. Autoshrink is available, but recommend that it is switched off and shrinking is done by maintenance jobs after backups.

23. Databases can be dbo use only (in development phase) (=restricted session); and can be in single user mode (when doing restores, etc.).

24. Truncate log on checkpoint is equivalent to noarchivelog mode in Oracle. Truncating the log on backup is equivalent to archivelog mode in Oracle.

25. On line redo logs are not archived, except on backup. They will therefore grow to much larger sizes than oracle On line redo logs.

26. By default, a database is created with one filegroup, named default. Microsoft recommend that filegroups should only be used for backup purposes. Performance should be handled by striping, even with tables and indexes.

27. Unlike our practice with Oracle on NT, but like our practice with Oracle on Sun: Microsoft recommend on line redo log files should be on separate physical disks, with separate disk controllers, from the datafiles. This is recommended both for performance and for fault tolerance. [p160]

28. During an online backup: cannot create or alter databases; create indexes; perform nonlogged operations such as bulk copy and writetext. These will be failed if attempted after backup is started, or cause backup to stop these are already running.

29. Three types of backup: full backup (=online backup); online redo log backup (=archive); and differential backup. The last type is completely new to SQL Server. It backs up just those blocks which have been modified since last full backup.

30. Create Index statements force the data and index filegroups to be backed up simultaneously.

31. There are no rollback segments in SQL Server. Rollback information is obtained from the online redo logs. This should improve performance.

32. SQL Server is read inconsistent. This will improve performance for some SQL jobs, but produce inconsistent results.

33. Standby databases can be up and available read-only on SQL Server. On Oracle7 they are unavailable.

34. SQL Server creates snapshot disk files, rather than snapshot log tables.

35. SQL Server has transactional replication, which replicates data by monitoring redo logs. block changes for replication are specifically marked in the redo logs. This is like snapshot replication, but with constant data copying. So it provides the same functionality as distributed database triggers, but with fewer potential problems.

36. A distribution database stores distribution history, and in transactional replication, also keeps the information culled from the redo logs for propagation.

37. SQL Server does not have parallel Server, although this may be possible via NT operating system.

38. SQL Server does not have sequences. Instead columns can be given the identity property.

39. By default, autocommits instantly, rollback unavailable.

40. With begin transaction...commit/rollback transaction statements, get commit, rollback, endpoint functionality. However, because SQL Server does not have rollback segments, modified rows in a transaction are locked. Users cannot even select from the entire table in most cases.