locked
Database Design RRS feed

  • Question

  • Hi everyone, I have a design question I hope you guys can help me.

    We have a requirement that is important to implement, I detail it as follows

    We have to process payments and charges that several companies capture from Point of Services (POS), we can have many companies for country, and right now we have to process 5 countries. So is a very importat amount of data.

    In order to avoid blockings and to improve IO, the client propose different scenarios, one is to create one database per country connected to a SAN por the storage.

    Another scenario is to create tables per country, but we have to create the same tables for every country, and the last scenario is to create schemas and create tables per country as well.

    All scenarios will be attached to the same SAN storage.

    What can be the best solution?

    Thanks for any advice.

    Friday, April 23, 2010 3:34 PM

Answers

  • the situation related with your conditions

    IO performance

    CPU performance

    query performance

    transaction counts etc.

     

    but simply if you will use one unique storage you can use partitioning

    as more complex if you want to optimize cpu and memory performance also you could use active-active cluster attached to same storage

    so you could distribute processes


    http://www.sdtslmn.com
    • Marked as answer by JAMC_SQL Friday, April 23, 2010 4:22 PM
    Friday, April 23, 2010 4:03 PM

All replies

  • IIUC, the quick answer is partitions. Though, a more detailed explanation of the setup and problem might help here.
    Friday, April 23, 2010 3:37 PM
    Answerer
  • the situation related with your conditions

    IO performance

    CPU performance

    query performance

    transaction counts etc.

     

    but simply if you will use one unique storage you can use partitioning

    as more complex if you want to optimize cpu and memory performance also you could use active-active cluster attached to same storage

    so you could distribute processes


    http://www.sdtslmn.com
    • Marked as answer by JAMC_SQL Friday, April 23, 2010 4:22 PM
    Friday, April 23, 2010 4:03 PM