Results 1 to 3 of 3
  1. #1
    Join Date
    Aug 2013
    Posts
    80

    Unanswered: Shared Nothing vs Shared Everything

    Hi All,
    DB2 DPF is running on Shared Nothing approach whereas Oracle RAC is using Shared Everything approach.
    First approach is has own dedicate memory and cpu for each partition node while later approach is shared all the resources for all the database files in the SAN Storage. I read as article that datawarehouse is more suitable for Shared nothing architecture and shared eveything is more towards OLTP environment.
    I know IBM has release Pureflex which to compete with Oracle RAC for this cause. Why shared everything can't scale as good as shared nothing ? I believe Oracle RAC one of the key advantage is Scalability. Can someone tell more details on this differences ? Any comments ?

    Thanks.

  2. #2
    Join Date
    Jun 2003
    Location
    Toronto, Canada
    Posts
    5,516
    Provided Answers: 1
    Quote Originally Posted by ckwan123 View Post
    Hi All,
    DB2 DPF is running on Shared Nothing approach whereas Oracle RAC is using Shared Everything approach.
    It's called "shared data", not "shared everything", because the database manager instances are separate while they access a single database.

    Quote Originally Posted by ckwan123 View Post
    I know IBM has release Pureflex which to compete with Oracle RAC
    PureFlex is just a hardware platform. I guess you mean DB2 pureScale.

    Quote Originally Posted by ckwan123 View Post
    I believe Oracle RAC one of the key advantage is Scalability.
    Good luck scaling beyond four nodes.
    ---
    "It does not work" is not a valid problem statement.

  3. #3
    Join Date
    May 2003
    Location
    USA
    Posts
    5,737
    Share nothing database architecture (DPF) is good for data warehouses since it allows parallelism for complex queries that access a lot of data each time. OLTP does not need (or want) each query to be parallelized, it just needs a lot of simultaneous queries to be executing at the same time, so that a platform like pureScale is a good solution (if the needed capacity exceeds a single database server).

    True share nothing database architectures existed back in the days before there were multiple CPU's and cores per server (mid 1990's). Each physical server typically only had one DPF partition. Now that we have many CPU cores per server, we have quasi-share nothing, which works quite well for DPF.
    M. A. Feldman
    IBM Certified DBA on DB2 for Linux, UNIX, and Windows
    IBM Certified DBA on DB2 for z/OS and OS/390

Posting Permissions

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