Re: Oracle or Sybase.

rowan.mccammon_at_macquarie.com.au (Rowan McCammon) wrote in <8s08iu$kbg$1_at_isdserv3.macbank>:

>Hi All,
>
>I'm not trying to start a war here ...

• snip --

I'd like to ask about "high availability". I was just researching this exact issue: Oracle vrs. Sybase. I'm coming from an OpenVMS perspective, so the argument became moot: Sybase canned OpenVMS support ending in version 11 - where version 12 is now out. However, I unearthed an interesting distinction between the two products.

A given Sybase client connects to a back end server, and from there, can access one of perhaps many databases. In the Oracle world, when you connect to a back end server, that dictates which physical database you have connected to.

But this gives Oracle an unusual advantage - multiple back end servers (instances) can connect to the same physical database (disk files) _concurrently_ - and they can be on different machines (see note 1). In the world of "high availability", you could take one machine down, and the database would still be available through the instance running on the other machine - only the users on the downed machine would need to re-connect to the surviving instance.

This is the approach I use to provide redundancy in high availability applications. It's called "Parallel Server". Sybase has their "High Availability Adaptive Server", but its structured differently (see above) and can't do this. If high availability really counts, Oracle's approach would appear to be superior.

ws

Note1: This feature requires that both boxes have direct access to the disks on which the database container files reside - often on a shared SCSI bus. Solaris and OpenVMS can do this for sure, NT can't. Oracle just preannounced  their intention to do this on specified Linux clusters. It vapor-ware right now.

-- 
<< What if there were no hypothetical questions? >>