Re: Partitioning Key

Collapse
This topic is closed.
X
X
 
  • Time
  • Show
Clear All
new posts
  • malapha@gmx.de

    Re: Partitioning Key

    On 11 Apr., 13:45, Serge Rielau <srie...@ca.ibm .comwrote:
    mala...@gmx.de wrote:
    Dear all,
    >
    I would like to hear your oppion about the follow partitioning keys.
    With the goal to have an evenly distributed dataset around all
    partitions we chose the following:
    >
    No
    1000001
    1000002
    1000003
    1000004
    1000005
    >
    This key works very well. Now we discuss a chance to the following
    >
    10000010
    10000020
    10000021
    10000022
    10000030
    10000040
    >
    Each Number is unique, but they may differ in the last column.
    Is this of any issue to the distribution of the partitions? Is the
    result the same?
    >
    The hashing for the partitioning key is not that trivial as to only take
    the integer value as is, so yes, if they values are unique you should
    always be fine.
    If, against all odds you observe skew you can always change weights in
    the DBPARTITION MAP which maps the 4000 hash values to your partitions.
    >
    Cheers
    Serge
    --
    Serge Rielau
    DB2 Solutions Development
    IBM Toronto Lab
    Thanks alot Serge. That helped..

    Regards
Working...