Re: benchmark kits for Oracle as well as SQL Server and others (Sybase, Postgres, MySQL, DB2)
Date: Thu, 6 Dec 2012 17:40:32 -0800
Message-ID: <CADsdiQgvNWvGrjEEVg5bfA-F2jY1ZROVBevtwjKJbmw9+DtA1g_at_mail.gmail.com>
When I say virtualization I mean virtualizing databases, as opposed to virtualizing machines. Virtualizing databases means having read only copies of datafiles that are shared between cloned databases. Each clone sees what appears to be a private copy of the database. Database changes made by a clone are only seen by that clone via a copy on write mechanism. In this case the only thing affected is I/O as everything else about the database stays the same.
Database Virtualization gives:
- Super fast provision of as many clones of a database as you want. Only takes a few seconds to stand up a fully functional, point-in-time clone of a database no matter the size. This means developers can make their own copies of production in seconds by themselves without asking their manager, their DBA, their storage admin etc etc. What took weeks in the past takes seconds now not only because of eliminating copying time of the production database but also and more importantly because all the time for requesting, discussing, processing and allocating resources for copies is eliminated. Because the resource cost is eliminated it also means that teams of developers can go from sharing one copy of production to each having their own private copy of databases. Now with a private copy of the database, a developer can change schema and meta data as fast as they want instead of waiting days or weeks of review time to check in changes to a shared development database.
- Storage savings as a new database copy only takes up space for private redo and temp. The original source backup is also compressed typically to 1/3 of their original size.With out Delphix, if I wanted to supply team of 100 developers each their own copy of a production database, then for every 1TB of production, it would cost 100TB of disk to support the copies. Copying 100TB of data would not only require massive amounts of storage but also be time consuming and copies would be out of date with production. With Delphix each 1TB of production requires about 0.3TB of storage and copies can be made in seconds by the developers themselves. New refreshes from production can be made in seconds, meaning the copies of production that developers use are in sync with production.
Here is some of the testing results working with swingbench
https://sites.google.com/site/oraclemonitor/ibm-and-delphix-performance-boosting-caching?pli=1
- Kyle
On Wed, Dec 5, 2012 at 9:28 PM, Purav Chovatia <puravc_at_gmail.com> wrote:
> Hi, We have been using hammerora but my experience is limited only to
> Oracle i.e. we have never tried benchmarking any other RDBMS using
> hammerora. We have found it to be good. Over the last 2 years there have
> been some enhancements as well, which are really cool.
> Its written in Tcl/Tk. Its free.
>
> For some time we have been researching on the impact of virtualization.
> Since you are doing benchmarking on virtualized platform, can you share
> some findings like % overhead, etc.?
> I already read that in your testing, virtualization is affecting only IO.
> Is it affecting throughput and latency both or just the latency? I would be
> particularly interested in latency impact.
>
> Thanks in advance.
>
>
> On Thu, Dec 6, 2012 at 2:08 AM, kyle Hailey <kylelf_at_gmail.com> wrote:
>
>>
>>
-- http://www.freelists.org/webpage/oracle-lReceived on Fri Dec 07 2012 - 02:40:32 CET