Re: 90GB table on Windows 2000

Billy Verreynne <vslabs_at_onwe.co.za> wrote in message news:<ao62bd$qfo$1_at_ctb-nnrp2.saix.net>...
> I agree with Nial and Howard. The stability of an operating system has less
> to do with the actual operating system itself, and much more to do with the
> ability of the person administrating it.

I will just add one real world scenario. I will put no comments on this, nor I'm pretending that this behaviour should has some impact in enterprise solutions.

We needed to do a stress testing of one very important database just migrated (in test environment) from 8 to 8i (EE 8.1.7 64-bit). Server was SunFire 6800 with 8 UltraSparc III processors and 8 gigs of RAM. Everything running in dedicated mode.
We simulate sessions performing repeating queries from three PIV with 512MB RAM (and gigs on HDDs available for virtual memory) using MS Windows 2000 both workstation and servers (in fact workstation and servers are technically exactly the same with exception of some CPU limits etc.).
We had started with 1000 concurrent sessions and by 200 increments went to about 6000 concurrent sessions. While reaching 6000 sessions, SunFire exchausted all physical memory together with about 10GB of virtual one (sort_are_size was rather big and many queries were sort oriented). Response times were not very good (nothing suprising), but machine was fully responsible. While using shell everything looks like usual, directory listing, ps, top etc worked almost instantly. On the other side all WIN2000 PCs (which basicaly were just sending requests), exchausted all of their physical memory. Any attempt for running more sessions ends with program crashed with exception (OS not DB one). WIN2000 couldn't use virtual memory for running testing programs and when physical memory became extinct, no more load on PCs could be handled. Even user interface was almost not responding. Opening "My Computer" took more than four minutes, using Ctrl-Alt-Del shows security (don't sure about right name) screen in about five minutes.
So we finally gave up and ended the test, mainly by killing test program instances, which cannot be closed properly. In just seconds SunFire fully recovered, freed virtual and physical memory. In no time response times went back to normal. PCs with WIN2000 spend few minutes heavily accesing the disks, trying to free virtual memory. However, clearing of allocated resources was not fully successful and PCs were restarted by their respective owners, because of not returning to former shape.

--
_________________________________________

Dusan Bolek, Ing.
Oracle team leader

Note: pagesflames_at_usa.net has been cancelled due to changes (maybe we
can call it an overture to bankruptcy) on that server. I'm still using
this email to prevent SPAM. Maybe one day I will change it and have a
proper mail even for news, but right now I can be reached by this
email.