Re: Data Centre Failover and Oracle Grid

From: DA Morgan <damorgan_at_x.washington.edu>
Date: Wed, 27 Apr 2005 11:35:10 -0700
Message-ID: <1114626671.698594@yasure>

Bernd Haug wrote:

> DA Morgan <damorgan_at_x.washington.edu> wrote:
>

>>Alex White wrote:
>>
>>>All I can say is that the connection between each data centre is a
>>>10Gb dark fibre, and the distance between sites is about 60 km.
>>
>>Then what you propose is technically impossible given any sum of
>>money your management will be willing to provide you.

>
>
> > 1.125 GIGABYTES PER SECOND (please don't mind my screaming)
> considering proto overhead are not enough? Or is it the expected latency
> when you have (nodes) - (switch) - (router) - (switch) - (wan endpoint)
> - (switch) - (router) - (switch) - (nodes) the obstacle?
>
> What is required to run RAC with multiple Datacenters? Infiniband with
> pretty short cables?

A day ago you didn't understand what the mem interconnect had to do with RAC. Today you are screaming GB/sec. The mem interconnect issues are not around bandwidth ... they are around latency.

There are many technical, and expensive technical, challenges around stretch clusters. To start with the mem interconnect MUST be a private network. So if that 1.125GB/sec is being used for anything else you can not use it ... or if you do you forfeit support from Oracle something you really don't want to do ever: Especially with RAC.

> Please excuse my newbie surprise[1] (I'm a Sysadmin (Solaris, Debian/RH
> Linux, Mac OS X) rather than a DBA), but I am trying to catch up on
> Oracle since I will be managing the systems beneath a RAC, probably on
> ASM storage, pretty soon.

> Right now I'm reading DB principles and 2dDBA in my "spare time" (yeah
> right, spare time), so you can see my current skill level. ;(
>
> lg, Bernd
>
> [1] read: horror

I admire your willingness to learn so decrease the volume and see if you can get the RAC docs from OTN. Also see if your SE can get you some of the various white papers published by the RAC Pack.

-- 
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace 'x' with 'u' to respond)

Received on Wed Apr 27 2005 - 13:35:10 CDT