Re: Log file partition
From: Geoff N. Hiten (SRDBA_at_Careerbuilder.com)
Date: 04/19/04
- Next message: Tom Moreau: "Re: How to prevent a lock escalation to a table level ?"
- Previous message: ShannonT: "Re: Scheduled DB Backup Never Executed, Why? Why?"
- In reply to: Steve Lin: "re: Log file partition"
- Next in thread: John Bandettini: "re: Log file partition"
- Messages sorted by: [ date ] [ thread ]
Date: Mon, 19 Apr 2004 18:48:09 -0400
Comments Inline
-- Geoff N. Hiten Microsoft SQL Server MVP Senior Database Administrator Careerbuilder.com I support the Professional Association for SQL Server www.sqlpass.org "Steve Lin" <lins@nospam.portptld.com> wrote in message news:eFvkrylJEHA.3628@TK2MSFTNGP12.phx.gbl... > I don't quite understand the reasons for the "standard 3 drive setup"? can > someone explain it a bit more for me? (who came up with such "standard"?) > if c is for OS > d is for sql excutables > e is for data, then my questions are: > > 1. in sql 2k, even specify program files being installed on d, there are > some files still being installed on c:\program files\mssql Yep. Most are system databases which are in simple mode anyway. The only things you might want to move are the tempdb devices. Once SQL is up and running you can change the default log and data file locations. > 2. if all data go to e (both mdf, ldf i assume), is that considered a good > performance (I/O) and fault tolerence (if the d drive gose bad, what happen > to trans log?) strategy? Again, the system databases with the exception of tempdb are now volume and simple recovery only. They are also fairly small and can be backed up daily. > > at my previous company, the set up is > C is for os and sql excutables > D is for logs > and F is for data and backup files. > c and d are one partitioned mirror drive. (for fault tolerance) > F drive is raid 5. Ooh, lousy performance AND all the eggs in one basket. Worst practices run rampant. > can someone tell me if the "standard 3 drive setup" is better than this set > up? and why? > Three drives are for fault tolerance and performance. All transactions are held up until the log entry is physically committed to disk. Also, log files are sequential writes and database files tend to be random writes. Disk head movement optimization algorithms can sometimes cause log writes to be delayed if data is on the same partition. The biggest reason to separate them is recovery. As long as you have a decent backup rotation and FULL recovery for the user databases, a three drive system will always allow you to recover all transactions up to the last moment before a hardware failure if it is a single partition failure. > Steve > > > > "Ed" <anonymous@discussions.microsoft.com> wrote in message > news:D3CAE674-8BFF-4CB0-A920-17C91D4A2093@microsoft.com... > > Diane, > > > > The standard 3 drive setup is to put the operating system on the C drive, > the application(SQL Server in this case) on the D drive and the data on the > E drive, in your case the .mdf files would be on E and your .ldf files would > be on another drive( F for example). > > > > Ed > > >
- Next message: Tom Moreau: "Re: How to prevent a lock escalation to a table level ?"
- Previous message: ShannonT: "Re: Scheduled DB Backup Never Executed, Why? Why?"
- In reply to: Steve Lin: "re: Log file partition"
- Next in thread: John Bandettini: "re: Log file partition"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
Loading
