Re: Db2, Oracle, SQL Server

From: DA Morgan <damorgan_at_x.washington.edu>
Date: Fri, 11 Feb 2005 15:50:27 -0800
Message-ID: <1108165657.122997@yasure>

Serge Rielau wrote:

> DA Morgan wrote:
>

>> Serge Rielau wrote:
>>
>>> What are you talking about? You can access all the data in DB2 from 
>>> any node.
>>
>>
>>
>> I may have jumped in on this thread without the correct context for the
>> original statement ... but I think the question related to a node loss.

>
> Indeed you did. The context is application transparency.
> Noons is claiming that queries in DPF need to contain node-specific
> language features (such aas a nodenum() function perhaps) to execute.
> Also if you would have read the context (or even the very post you
> replied to) you woudl have read my argument that no one the F..K (what
> doe sthat stand for? is it a technical term?) cares what happens when
> only 9 of 10 nodes are there because that is much or little an issue as
> what happens when a RAC cluster remasters.
> In RAC you remaster and live without the node. In DB2 you restart the
> friggin node. It is my no means hard nor time consuming.
> And now I will stop posting to this thread because it is loosing it's
> technical content. Instead we are down to simply re-hashing the same
> words, for which I have neither time nor patience.
>
> Cheers
> Serge

You are correct AFAIK that in DB2 there is no need to write nod-specific language. That is reserved for SQL Server's federated design.

But you jumping into a lifeboat I interpret to mean that I am correct, even if not tracking with the thread, that if you lose a node in DB2 you are toast for some portion of your data.

BTW: The demo I use of shared-nothing vs. shared everything shows four nodes. A-E, F-K, L-Q, R-Z. What happens if node 2 is lost is self-evident from the data it contains eh.

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

Received on Fri Feb 11 2005 - 17:50:27 CST