Re: oracle9i preformance on WIndows Server

Sybrand Bakker <sybrandb_at_hccnet.nl> wrote in message news:<8ljrd0hd126prmh6kqgcpjgagca0qgqv5o_at_4ax.com>...
> On 26 Jun 2004 10:59:01 -0700, premmehrotra_at_hotmail.com (Prem K
> Mehrotra) wrote:
>
> >My managemnt wants to use Windows servers instead of UNIX servers to
> >save costs. LINUX is not an option because my data center does not support
> >it. We are looking into buying HP's ProLiant G2 BL20p model of servers. It has
> >2 CPU's. We will get maximum memory of 8GB.
> >
> >We may get 1 or 2 internal drives. My database is not that big ~10GB.
> >However, 100 of users will be concurrently accessing it. Also, it will
> >be used for OLTP as well as for reporting.
> >
> >I have not worked with Windows server earlier. Are there benchmarks avaiable
> >for WIindows server vs HP UNIx servers. I have nothing against using Windows
> >Servers, just want to make sure performnace will not be an issue.
> >
> >
> >Thank you.
> >
> >Prem
>
>
> Google is your friend as usual. Recently (IIRC Howard Rogers)
> conducted a comparative study between Linux and Windoze.
> Needless to say Linux was much faster.
>
> Note also you need Windows Advanced Server to make use of any memory
> beyond 4 G at all, and Dead Connection Detection will not work, and
> prespawned servers will also not work.
> You may be in for some unpleasant suprises.
> By design the Windows O/S has limited scalability unlike Unix.
>
> Please make sure you have your management sign a letter you aren't
> responsible for any performance issues, which result from the platform
> swap.

"Dead Connection Detection will not work" Sybrand speaketh the truth here.

nothing like a user that is blocked but is holding locks hitting <ctrl>+<alt>+<del>, logging back into the app, and repeating what they just did, slamming right into the locks held by their now zombied session.

logjam.

be prepared to do some database babysitting, and rolling a routine to (ora)kill zombies.

this is without a doubt the number one PITA running on win32.

-bdbafh Received on Sat Jun 26 2004 - 21:04:00 CDT