Joe Smith wrote:
>
> 5. If you are not using Oracle in client server mode (your users sign on to
> the HP then access Oracle) then you will not see the following problems.
> The Oracle listener spawns system processes for each remote user (Oracle
> 7.1.6).  Since they are system processes,  their priority is linear, ie
> they are never decremented when they exceed their quantum.  This means
> that, since they default to the CS queue, a large query or run-away process
> can monopolize the system and cause you to have to shut down and reboot.
> The listener process is expected to be changed to a batch process spawning
> normal non-system processes in 7.3, so this problem should be fixed. We
> retuned our queues to circumvent this problem for now.
>
> 6. Dead Connection Detection (DCD) does not work correctly as of yet.  This
> kills off orphaned processes when a user does a hard disconnect, such as
> rebooting the pc.  This typically throws the process into a cpu loop and
> brings the system to its knees.  This problem also exists on the Unix
> platform, so you can't get away from this in client server mode.
>
While I cannot throw in additional Oracle comments, the 2 above
scenarios
sound like the <plug> Workload Manager product </plug> would complement
Oracle quite nicely. Unfortunately the Workload manager is not
'complimentary '.

Goetz