Re: Larry Ellison comments on Microsoft's benchmark

Which other published TPC-C results have used mirrored disk for logs and data? Which other TPC-C results have priced in a High Availability software package (like Veritas, MSCS, Sun Cluster, AIX HACMP or HP ServiceGuard) as well as the database scripts for HA? At least DB2 ships HA scripts as part of the base product.

I suspect that mirrored disk are used for only some of the other TPC-C results, and that HA software packages are part of no one's TPC-C pricing. However, I would like to know if I am wrong. As to the number one result being meaningless, I disagree:
- the performance was achieved
- pricing may change if you built serious HA into the benchmark, but it would change for other results as well (in some cases more drastically for the results that lack mirrored disk for data)

Ivana Humpalot wrote:

> Thank you for your honest and straightforward answer.
>
> This basically means the TPC-C benchmark is not extremely useful.
>
> Although IBM's result tops the charts, the IBM system as used
> in the benchmark test may not be as reliable as some other vendor's
> system which may be further down the TPC-C chart.
>
> Unlike in TPC-C, in real life availability is important, so the
> position in the TPC-C chart is meaningless.
>
> "Blair Kenneth Adamache" <adamache_at_ca.ibm.com> wrote in message
> news:39629B7D.714D7427_at_ca.ibm.com...
> > The published benchmark does not include high availability machines or
> > Microsoft Cluster Server (MSCS). If you wanted to back up each machine,
> > you'd need hot standby machines (doubling the number of CPU's and RAM,
> > but not increasing the disk), mutual takeover machines running some other
> > workload, or have the machines back each other up (which is very common
> > for DB2 EEE customers). We didn't publish with true HA because no one
> > else does, and we wanted a decent price metric. We already have more
> > redundancy than other TPC publications with our UPS and mirrored data
> > disks.
> >
> > The published configuration will give correct answers only. If a node
> > goes down, and MSCS is not used to back it up, SQL that accesses data on
> > that node will return with a negative SQLCODE. SQL that does not touch
> > that node will succeed and give correct results.
> >
> > Ivana Humpalot wrote:
> >
> > > Blair,
> > >
> > > My question was not about the hard disk. My question was what
> > > will happen if one of the MACHINES fail.
> > >
> > > You appear to be avoiding the question, though that may not
> > > have been your intent.
> > >
> > > If one of the MACHINES fail, will the queries still give the
> > > right answer?
> > >
> > > Will the configuration as tested, without using ANY additional
> > > hardware, continue to give correct answers even if a machine
> > > fails?
> > >
> > > Please clarify. I am sure a lot of people are interested in
> > > the answer to this critical question.
> > >
> > > "Blair Kenneth Adamache" <adamache_at_ca.ibm.com> wrote:
> > > >
> > > > Unlike many TPC results, DB2 results always publish with
> > > > mirrored disk. We mirrored the data disks with RAID 1E,
> > > > and the log disks with RAID 5. DB2 also used a UPS
> > > > (Uninterrupted Power Supply) - these are all features
> > > > that raised the price, but made the result more real-world.
> > > >
> > > > Ivana Humpalot wrote:
> > > >
> > > > > Blair,
> > > > >
> > > > > Can you answer a question that I believe is highly relevant and
> > > > > important:
> > > > >
> > > > > Does Larry Ellison's comments apply to the top TPC-C result (DB2
> > > > > on NT)?
> > > > >
> > > > > In other words, if one of the machines fail, will DB2 still give
> > > > > me the right results? Or will queries fail or give wrong answers?
> > > > >
> > > > > What I want to know is whether the configuration as tested can be
> > > > > used in real life without using any additional hardware.
> > > > >
> >
Received on Wed Jul 05 2000 - 00:00:00 CDT