Results 1 to 7 of 7

Thread: RAID Confusion

  1. #1
    Join Date
    Oct 2013
    Posts
    6

    Red face RAID Confusion

    Now, there probably is a proper forum to place this in, but I'm not entirely sure where it should go. I don't think anyone will be too bothered by me posting it in this forum too much (better than somewhere completely inappropriate).

    You see, I though that I understood RAID. Well, I sort of do. I get the principles behind how Raid 0,1,5 and 1+0 work at the very least. However, I thought you could only use one raid at a time. Recently, however, I've seen people talking about separate RAIDs for the OS, DBMS software, Database files, transaction logs, and backup files. Seems to me having separate RAID for each one of these things would require a lot of HDDs or SDDs (maybe it doesn't. This stuff is still new to me). I'm curious as to whether this is the common practice when setting up a database server environment.

    Should I actually have completely separate RAID models for my OS, DBMS software, transaction files and backup logs? Can I mix and match to some extent, i.e. have two or three of them share the same RAID. Are each of things actually necessarily associated with separate HDDs/SDDs from one another?

    It probably seems like a very stupid question, but I concede to not really having a clue what I'm talking about. I have to start somewhere

  2. #2
    Join Date
    Feb 2004
    Location
    In front of the computer
    Posts
    15,579
    RAID exists fundamentally at the "spindle" level, the smallest addressable disk device. That is the whole point of the RAID name/acronym (Redundant Array of Inexpensive Disk). For many technologies the base spindle size is around 4 Gb, so it takes a small hoard of devices to make a usable storage array.

    Lets assume that you want to build a 500 Gb system even though that is on the small side today. Quick math shows that you will need 125 drives at 4 Gb each.

    You can divvy up the drives in any way that makes sense to you, and can apply RAID to any group of drives as long as the drive geometry works. An example that wouldn't work is to apply RAID 1 + 0 to an array of seven drives... That doesn't fly because 1 + 0 requires an even number of spindles.

    At least based on my experience, getting too fancy with the RAID architecture doesn't make a measurable difference for our systems because other components in the storage system become bottlenecks long before the RAID striping gets to be the weakest link. It is critically important to know where the "weak spot" for performance is in your system instead of just assuming that it is where you expect (or have been taught) that it will be.

    -PatP
    In theory, theory and practice are identical. In practice, theory and practice are unrelated.

  3. #3
    Join Date
    Sep 2009
    Location
    Ontario
    Posts
    1,057
    We use multiple RAID logical drives.
    Generally we use 2 30gb SSD in RAID1 for the operating system, and then build additional RAID1/10 logical drives for the data.
    You need to do a little statistical analysis to determine the ratio of reads to writes.
    RAID1 performs better with higher read ratios, since the controller has a choice of more than one spindle to retrieve data from. Writes are slower (than RAID5) because the data has to be written in two places.

  4. #4
    Join Date
    Oct 2013
    Posts
    6
    That's interesting Kitman. Would you say that 2 30gb drives in a RAID 1 configuration would also be acceptable for the transaction logs, or should you need more/less capacity compared to the OS?

    While I was thinking about this, I realised that there are two more things I don't understand.

    1. Is there a distinction between transaction logs and redo logs?

    2. Is there a difference between a "backup server" and "backup set"?
    Last edited by Meloda; 10-22-13 at 12:25.

  5. #5
    Join Date
    Oct 2013
    Posts
    6
    Just to see if I'm catching on, would this picture constitute a whole DBMS?

    Click image for larger version. 

Name:	DBMS.PNG 
Views:	2 
Size:	20.6 KB 
ID:	14870
    Last edited by Meloda; 10-22-13 at 12:29.

  6. #6
    Join Date
    Sep 2009
    Location
    Ontario
    Posts
    1,057
    The size of the transaction logs depends upon the number of transactions and the size of each transaction, and the time frame between purges.
    A redo-log is a file containing all the changes to a database since the last backup. It is used to bring a restored database back to current without re-keying etc.
    A transaction log (probably) lasts longer, is more of a history file, and contains copies of transactions posted to the database. Things like correcting the spelling of someone's name might not be in the transaction log, but would be in the redo-log.
    To me a backup set is a system backup consisting of more than one piece of media; there should be two dvd's in last night's backup set.
    A backup server, could be a machine to which a database is backed up to, ie the machine is the backup media, or it could be a clone of the original designed to replace the original in case the primary machine fails.

  7. #7
    Join Date
    Oct 2013
    Posts
    6
    Awesome explanation. Thanks!

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •