Re: Sybase vs Oracle - which is better?

lburns4770_at_my-dejanews.com wrote:
>
> One major difference between the DBMSs seems to be the way they use the
> operating system's shared memory. Sybase appears to do everything inside of
> one Unix process (the dataserver process),

	You can define from 1 to N engines with Sybase, with each engine being
	an OS process.

> while Oracle and Informix spawn
> multiple Unix processes to perform tasks. For this reason, Sybase requires
> you to define a huge chunk of shared memory on the Unix platform (2/3 to 3/4

	You can allocate as much (or as little) shared memory to Sybase as you
	want it to use.

> of the total physical memory), and allocate most of it to Sybase. Oracle and
> Informix require a shared memory pool that is only 20-25% of the total
> memory.
>
> Incidently, there doesn't seem to be any useable documentation anywhere on
> what settings to use for SHMMAX, although you can hose your system really
> badly if it is set too high or too low.

	In what sense? This would be OS dependant, but at least on Solaris
	you can set the value to anything. However, an application trying to
	grab more than is physically available is a different matter.

> I don't know why Sybase doesn't
> address this issue in any of its documentation.

	It doesn't need to. All you need do is to set it to what you want
	it to use. Or configure Sybase to use some portion of the available
	amount.

> Also, does anyone know the history of why Sybase is architected like this?

	It more efficient (at least on Solaris) to use one larger chunk rather
	than several smaller chunks. (And its also better to allocate one
	contiguous chunk rather than have fragmented chunks.)

-am Received on Wed Nov 18 1998 - 00:00:00 CST