Re: Serious article on comparison between MS SQL Server 2005 Yukon and Oracle 10g

Paul wrote:

> Daniel,
>
> I believe that you misunderstood that particular paragraph.

I belive otherwise.

> They specifically state that a cold failover of SQL server is
> sometimes
> faster than a TAF Reconfiguration of a RAC cluster,
> which in itself is rather surprising considering that RAC is
> advertised
> as a superior high availability solution.

First off what they wrote is pure nonsense. I have never seen a TAF failover that took more than two seconds with most being sub-second. But even if it were true ... which it clearly is not ... the correct comparison would be cold failover in SQL Server in comparison with cold failover in DataGuard.

What you are proposing is that I compare the time it takes to insert 1 record into a table on SQL Server to inserting one record on Oracle except on Oracle you have turned on archive log mode and the table has six triggers on it. The authors aren't comparing apples with apples. And, as I pointed out above, even if they were their conclusion is flawed.

> There is also no statement of precise reconfiguration time for RAC
> only that
> HP Sun Cluster reconfiguration time is 10-30 seconds.

Which is pure drivel. What reconfiguration time?

> I have some experience with RAC and the reconfiguration process is
> not immediate, to say the least. It of course depends on the size of
> the redo logs and the extent of recovery which needs to be performed.
> If the database has normal size redo logs (mine has redo logs of
> 200MB) and a heavy transaction load there will be a significant delay
> for reconfiguration when one of the nodes fails.

I am not aware of any excuse for what you are reporting.

> So I guess your database has redo logs of 10MB, no transaction load or
> some very fast hardware if reconfiguration takes 2 seconds.

Very fast hardware and certainly not 200MB redo logs. But there may be other things at work here. One is whether your operating system and or NIC configurations attempt to reconnect for some period of time before failing.

I can easily make RAC behave as you describe by configuring the system to attempt to re-establish a connection before failing. The best configurations are the dumbest ones. Check it out.

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