This discussion is archived
5 Replies Latest reply: Apr 16, 2012 10:42 PM by BillyVerreynne RSS

ASM will improve Disk I/O rate?

OraFighter Newbie
Currently Being Moderated
If I am using ASM for my database , is it any performance improvement I can expect ? -- in disk I/O rate,

and Is it compulsory to have multiple Physical disk itself for high disk i/o rate ?

or if I partition a single physical disk into multiple logical disk . I can make more i/o rate ?
  • 1. Re: ASM will improve Disk I/O rate?
    KuljeetPalSingh Guru
    Currently Being Moderated
    one of the key feature of ASM - Delivers predictable performance, availability and scalability

    Automatic Storage Manager spreads its files across all available storage for optimal performance. ASM enhances the concept of striping by providing additional flexibility for changes in the disk configuration. Spreading the workload across as many disks as possible results in less contention and high I/O throughput.


    http://www.oracle.com/technetwork/database/asmwp-130989.pdf

    Thanks
    Kuljeet Pal Singh
  • 2. Re: ASM will improve Disk I/O rate?
    OraFighter Newbie
    Currently Being Moderated
    Kuljeet Pal Singh wrote:
    Spreading the workload across as many disks as possible results in less contention and high I/O throughput.
    workload across as many disks means Logical disk is enough or we need separate Physical disk itself ..?
  • 3. Re: ASM will improve Disk I/O rate?
    KuljeetPalSingh Guru
    Currently Being Moderated
    You can create an ASM disk group using one of the following storage resources:

    Raw disk partition—A raw partition can be the entire disk drive or a section of a disk drive. However, the ASM disk cannot be in a partition that includes the partition table because the partition table can be overwritten.

    Logical unit numbers (LUNs)—Using hardware RAID functionality to create LUNs is a recommended approach. Storage hardware RAID 0+1 or RAID5, and other RAID configurations, can be provided to ASM as ASM disks.

    Raw logical volumes (LVM)—LVMs are supported in less complicated configurations where an LVM is mapped to a LUN, or an LVM uses disks or raw partitions. LVM configurations are not recommended by Oracle because they create a duplication of functionality. Oracle also does not recommended using LVMs for mirroring because ASM already provides mirroring.

    NFS files—ASM supports NFS files as ASM disks. Oracle Database has built-in support for the network file system (NFS) and does not depend on OS support for NFS. Although NFS and ASM have overlapping functionality, ASM can load balance or mirror across NFS files.



    http://docs.oracle.com/cd/B28359_01/server.111/b31107/asmprepare.htm
  • 4. Re: ASM will improve Disk I/O rate?
    onedbguru Pro
    Currently Being Moderated
    Hmm, lets see... ASM provides parallel read/write/redundancy and you want a disk (as an example) with 2 partitions 500G/each .. Single set of read heads, Single set of platter(s) If you really think about what is a logical disk (in your example a single device with multiple partitions) are you sure you want to go there? What happens when that single spindle dies... POOF!!! there goes your database - all of it.

    I have moved MANY databases from file system to ASM - typical performance improvement (with a PROPERLY configured ASM) anywhere between 10-20%. Some much higher. YMMV.

    Having dealt with NFS - it is NOT a configuration I would consider. (used it for a test and dumped it!!)
  • 5. Re: ASM will improve Disk I/O rate?
    BillyVerreynne Oracle ACE
    Currently Being Moderated
    OraFighter wrote:
    If I am using ASM for my database , is it any performance improvement I can expect ? -- in disk I/O rate,
    Simple answer. No.

    ASM is not a magic wand to wave that suddenly makes a storage layer faster. A storage layer has a physical speed limit. You cannot exceed that.

    ASM provides you with the means of using that storage layer optimally and managing it effectively.
    and Is it compulsory to have multiple Physical disk itself for high disk i/o rate ?
    Not a proper question as "high" I/O rates are dependent on a number of factors. One of which is the number of physical disks (as that decreases I/O latency) and the actual physical read/write/seek times of the disk (as that increases raw I/O call performance).

    But there are a number of other equally important factors too - ranging from how the disks are used (RAID levels), how the storage cache is configured, to the storage protocol used and the type of I/O fabric layer used.
    or if I partition a single physical disk into multiple logical disk . I can make more i/o rate ?
    No.

Legend

  • Correct Answers - 10 points
  • Helpful Answers - 5 points