Re: Larry Ellison comments on Microsoft's benchmark

From: Blair Kenneth Adamache <adamache_at_ca.ibm.com>
Date: 2000/07/06
Message-ID: <39641169.5C934DDC@ca.ibm.com>#1/1

Ivana, has OPS been used to publish a recent TPC-C benchmark? The top ten only seems to have Oracle 8i and Oracle 7 results. The only OPS result I can find is from 12/98, somewhere around 37th place (shown below).

If could beg a small indulgence, Ms. Humpalot, does OPS run into some scaling difficulties because of shared disk? My view of HA on shared nothing is that if you have 12 nodes, each node can be backed up in a mutual takeover configuration by 1-11 other nodes (depending on whether you cluster for HA pair wise, or use something more complex). No need for a distributed lock manager to resolve who owns the disk. NODE A dies - NODE B takes over. In shared disk, NODE A dies, and NODE B fights with other nodes using that disk to save the work that NODE A was doing. 12 times as likely to fail? Perhaps, but far less likely to run into contention for shared devices, as nothing gets shared.

I should also state my bias, and point out that Blair Kenneth Adamache is not my real name (actually it is, but I am trying to get my coworkers to address me as "Ensign").

The OPS result I could find:

Compaq ProLiant 6500-PDC/01000
TPC-C Rev 3.5
Report Date: 12/23/98

Total System Cost                                       TPC-C

Throughput
Price/Performance
1692980.00 US $
33935.90
49.88

Database Manager: Oracle 8 Parallel Server v 8.05 Operating System: Windows NT 4.0

  CPU:            Intel Pentium II Xeon 400 MHz
  # of CPUs:            16
  Cluster:             Y

Ivana Humpalot wrote:

> "Garfield A. Lewis" <galewis_at_ca.ibm.com> wrote in message:
> > >
> > > 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
> >
> > Please..., let us know of any other TPC-C benchmark that includes HA...
> >
> > >
> > > 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.
> >
> > We do not say that HA is not important but whatever costs are associated
 with
> > it will be at least equally applied across any real world setup based on
 the
> > benchmark costs.
> >
>
> Oracle Parallel Server is designed for high availability.
>
> In Microsoft's and IBM's configuration, the database is split among
> many different machines. If any one of those machines fail the database
> becomes unavailable. This means a configuration that uses 12 machines
> is 12 times LESS reliable as a configuration that only uses one
> machine.
>
> In Oracle Parallel Server, a configuration that uses 12 machines is
> 12 times MORE reliable than a configuration that only uses one machine.
> However it should be noted that the disk is still a single point of
> failure. But this is still 12 times more reliable than an IBM or
> Microsoft configuration consisting of 12 machines.
>
> This is why I said that 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. Unlike Oracle, IBM's performance comes at the
> cost of reliability. (Unless additional hardware is used to improve
> reliability, which dramatically drives up costs.)
Received on Thu Jul 06 2000 - 00:00:00 CDT