Latest TPC bencmarks killing SQl Server

After Microsoft's doing so well in the TPC raw performance bechmark for so
long, Oracle appears to have finally leapfrogged SQL Server on both
Clustered and Non-Clustered configurations. In fact, not only has Oracle
leapfrogged them but literally blew SQL Server away.
What is REALLY interesting about this is the fact that the number-one
performance for the clustered configuration is both considerably higher than
Microsoft and WITH BETTER PRICE PERFORMANCE, which has always been
Microsoft's strength.
This appears to be a function of using an Open Source OS (Linux).
I still generally prefer SQL Server over Oracle (easier to tune and
maintain) but have been troubled by Microsoft's trend AWAY from commodity
pricing (that they pioneered at the high end) at the same time Enterprise
providers have been trending toward it. I'm wondering if this latest
development might single an inflection point where Microsoft begins to
rethink ever-upward pricing and once again return to a more aggresive
position in this area.It's a constant battle to keep the numbers up. If MS posts something Oracle
or DB2 will better it and then MS will have their turn. It's almost
pointless now that the numbers are so far out of what systems need now a
days anyway. The hardware has caught up with the rest of the world in this
respect and it's only getting better.
--
Andrew J. Kelly
SQL Server MVP
"nospam" <nfr@.nospam.com> wrote in message
news:%23%236UwKmzDHA.2000@.TK2MSFTNGP11.phx.gbl...
> After Microsoft's doing so well in the TPC raw performance bechmark for so
> long, Oracle appears to have finally leapfrogged SQL Server on both
> Clustered and Non-Clustered configurations. In fact, not only has Oracle
> leapfrogged them but literally blew SQL Server away.
> What is REALLY interesting about this is the fact that the number-one
> performance for the clustered configuration is both considerably higher
than
> Microsoft and WITH BETTER PRICE PERFORMANCE, which has always been
> Microsoft's strength.
> This appears to be a function of using an Open Source OS (Linux).
> I still generally prefer SQL Server over Oracle (easier to tune and
> maintain) but have been troubled by Microsoft's trend AWAY from commodity
> pricing (that they pioneered at the high end) at the same time Enterprise
> providers have been trending toward it. I'm wondering if this latest
> development might single an inflection point where Microsoft begins to
> rethink ever-upward pricing and once again return to a more aggresive
> position in this area.
>|||As far as non-clustered goes, something you need to take into account is
that the top Oracle NC result is on a box that has 1tb of main memory. The
top MS SQL NC result is on virtually the same box, but only has half the
memory. It's reasonable to assume that the top MS SQL NC result is probably
similar to the top Oracle NC result if they were run on the same hardware.
As far as the use of Linux is concerned, the O/S cost is only a small
fraction of the overall cost of that benchmark. Download the full disclosure
statement & I'd suggest you'll see that the cost achieved by Oracle is due
to commodity pricing of grid components & the cost saving from use of the
Linux O/S is a relatively small contributor to the overall cost saving on
Oracle's C result.
The last thing to take into account is that Oracle's result is based on 10g.
It will be interesting to see what Yukon can do.
Regards,
Greg Linwood
SQL Server MVP
"nospam" <nfr@.nospam.com> wrote in message
news:%23%236UwKmzDHA.2000@.TK2MSFTNGP11.phx.gbl...
> After Microsoft's doing so well in the TPC raw performance bechmark for so
> long, Oracle appears to have finally leapfrogged SQL Server on both
> Clustered and Non-Clustered configurations. In fact, not only has Oracle
> leapfrogged them but literally blew SQL Server away.
> What is REALLY interesting about this is the fact that the number-one
> performance for the clustered configuration is both considerably higher
than
> Microsoft and WITH BETTER PRICE PERFORMANCE, which has always been
> Microsoft's strength.
> This appears to be a function of using an Open Source OS (Linux).
> I still generally prefer SQL Server over Oracle (easier to tune and
> maintain) but have been troubled by Microsoft's trend AWAY from commodity
> pricing (that they pioneered at the high end) at the same time Enterprise
> providers have been trending toward it. I'm wondering if this latest
> development might single an inflection point where Microsoft begins to
> rethink ever-upward pricing and once again return to a more aggresive
> position in this area.
>|||Agree with all your points.
I would actually like to see one of these top-end benchmarks at least run on
as identical hardware as possible, such as the SuperDome, with the variants
being the database, TP Monitor, and OS. That is an integrated MS solution
vs. a non-MS "best-of-breed" approach.
IOW, I'd like to see the HW held constant and look for performance advantage
in software platform configurations.
"Greg Linwood" <g_linwoodQhotmail.com> wrote in message
news:ebPCK8mzDHA.1524@.TK2MSFTNGP10.phx.gbl...
> As far as non-clustered goes, something you need to take into account is
> that the top Oracle NC result is on a box that has 1tb of main memory. The
> top MS SQL NC result is on virtually the same box, but only has half the
> memory. It's reasonable to assume that the top MS SQL NC result is
probably
> similar to the top Oracle NC result if they were run on the same hardware.
> As far as the use of Linux is concerned, the O/S cost is only a small
> fraction of the overall cost of that benchmark. Download the full
disclosure
> statement & I'd suggest you'll see that the cost achieved by Oracle is due
> to commodity pricing of grid components & the cost saving from use of the
> Linux O/S is a relatively small contributor to the overall cost saving on
> Oracle's C result.
> The last thing to take into account is that Oracle's result is based on
10g.
> It will be interesting to see what Yukon can do.
> Regards,
> Greg Linwood
> SQL Server MVP
> "nospam" <nfr@.nospam.com> wrote in message
> news:%23%236UwKmzDHA.2000@.TK2MSFTNGP11.phx.gbl...
> > After Microsoft's doing so well in the TPC raw performance bechmark for
so
> > long, Oracle appears to have finally leapfrogged SQL Server on both
> > Clustered and Non-Clustered configurations. In fact, not only has Oracle
> > leapfrogged them but literally blew SQL Server away.
> >
> > What is REALLY interesting about this is the fact that the number-one
> > performance for the clustered configuration is both considerably higher
> than
> > Microsoft and WITH BETTER PRICE PERFORMANCE, which has always been
> > Microsoft's strength.
> >
> > This appears to be a function of using an Open Source OS (Linux).
> >
> > I still generally prefer SQL Server over Oracle (easier to tune and
> > maintain) but have been troubled by Microsoft's trend AWAY from
commodity
> > pricing (that they pioneered at the high end) at the same time
Enterprise
> > providers have been trending toward it. I'm wondering if this latest
> > development might single an inflection point where Microsoft begins to
> > rethink ever-upward pricing and once again return to a more aggresive
> > position in this area.
> >
> >
>|||You're not likely to see hardware held constant b/c these benchmarks are
usually released by hardware vendors - they'll always want to show off how
fast their latest "babies" are. (c:
Regards,
Greg Linwood
SQL Server MVP
"nospam" <nfr@.nospam.com> wrote in message
news:%23D$cbQnzDHA.3416@.tk2msftngp13.phx.gbl...
> Agree with all your points.
> I would actually like to see one of these top-end benchmarks at least run
on
> as identical hardware as possible, such as the SuperDome, with the
variants
> being the database, TP Monitor, and OS. That is an integrated MS solution
> vs. a non-MS "best-of-breed" approach.
> IOW, I'd like to see the HW held constant and look for performance
advantage
> in software platform configurations.
> "Greg Linwood" <g_linwoodQhotmail.com> wrote in message
> news:ebPCK8mzDHA.1524@.TK2MSFTNGP10.phx.gbl...
> > As far as non-clustered goes, something you need to take into account is
> > that the top Oracle NC result is on a box that has 1tb of main memory.
The
> > top MS SQL NC result is on virtually the same box, but only has half the
> > memory. It's reasonable to assume that the top MS SQL NC result is
> probably
> > similar to the top Oracle NC result if they were run on the same
hardware.
> >
> > As far as the use of Linux is concerned, the O/S cost is only a small
> > fraction of the overall cost of that benchmark. Download the full
> disclosure
> > statement & I'd suggest you'll see that the cost achieved by Oracle is
due
> > to commodity pricing of grid components & the cost saving from use of
the
> > Linux O/S is a relatively small contributor to the overall cost saving
on
> > Oracle's C result.
> >
> > The last thing to take into account is that Oracle's result is based on
> 10g.
> > It will be interesting to see what Yukon can do.
> >
> > Regards,
> > Greg Linwood
> > SQL Server MVP
> >
> > "nospam" <nfr@.nospam.com> wrote in message
> > news:%23%236UwKmzDHA.2000@.TK2MSFTNGP11.phx.gbl...
> > > After Microsoft's doing so well in the TPC raw performance bechmark
for
> so
> > > long, Oracle appears to have finally leapfrogged SQL Server on both
> > > Clustered and Non-Clustered configurations. In fact, not only has
Oracle
> > > leapfrogged them but literally blew SQL Server away.
> > >
> > > What is REALLY interesting about this is the fact that the number-one
> > > performance for the clustered configuration is both considerably
higher
> > than
> > > Microsoft and WITH BETTER PRICE PERFORMANCE, which has always been
> > > Microsoft's strength.
> > >
> > > This appears to be a function of using an Open Source OS (Linux).
> > >
> > > I still generally prefer SQL Server over Oracle (easier to tune and
> > > maintain) but have been troubled by Microsoft's trend AWAY from
> commodity
> > > pricing (that they pioneered at the high end) at the same time
> Enterprise
> > > providers have been trending toward it. I'm wondering if this latest
> > > development might single an inflection point where Microsoft begins to
> > > rethink ever-upward pricing and once again return to a more aggresive
> > > position in this area.
> > >
> > >
> >
> >
>|||3 out of the top 4 benchmark's are on the HP Integrity Superdome, which
theoretically could have been configured identically. The most recent number
1 slot goes to the HP Integrity rx5670, which was a clustered solution.
It's intereesting there is not a single Java-based TP-monitor in the bunch.
It appears the battle is between BEA Tuxedo and COM+.
"Greg Linwood" <g_linwoodQhotmail.com> wrote in message
news:%23ep$xYnzDHA.2160@.TK2MSFTNGP12.phx.gbl...
> You're not likely to see hardware held constant b/c these benchmarks are
> usually released by hardware vendors - they'll always want to show off how
> fast their latest "babies" are. (c:
> Regards,
> Greg Linwood
> SQL Server MVP
> "nospam" <nfr@.nospam.com> wrote in message
> news:%23D$cbQnzDHA.3416@.tk2msftngp13.phx.gbl...
> > Agree with all your points.
> >
> > I would actually like to see one of these top-end benchmarks at least
run
> on
> > as identical hardware as possible, such as the SuperDome, with the
> variants
> > being the database, TP Monitor, and OS. That is an integrated MS
solution
> > vs. a non-MS "best-of-breed" approach.
> >
> > IOW, I'd like to see the HW held constant and look for performance
> advantage
> > in software platform configurations.
> >
> > "Greg Linwood" <g_linwoodQhotmail.com> wrote in message
> > news:ebPCK8mzDHA.1524@.TK2MSFTNGP10.phx.gbl...
> > > As far as non-clustered goes, something you need to take into account
is
> > > that the top Oracle NC result is on a box that has 1tb of main memory.
> The
> > > top MS SQL NC result is on virtually the same box, but only has half
the
> > > memory. It's reasonable to assume that the top MS SQL NC result is
> > probably
> > > similar to the top Oracle NC result if they were run on the same
> hardware.
> > >
> > > As far as the use of Linux is concerned, the O/S cost is only a small
> > > fraction of the overall cost of that benchmark. Download the full
> > disclosure
> > > statement & I'd suggest you'll see that the cost achieved by Oracle is
> due
> > > to commodity pricing of grid components & the cost saving from use of
> the
> > > Linux O/S is a relatively small contributor to the overall cost saving
> on
> > > Oracle's C result.
> > >
> > > The last thing to take into account is that Oracle's result is based
on
> > 10g.
> > > It will be interesting to see what Yukon can do.
> > >
> > > Regards,
> > > Greg Linwood
> > > SQL Server MVP
> > >
> > > "nospam" <nfr@.nospam.com> wrote in message
> > > news:%23%236UwKmzDHA.2000@.TK2MSFTNGP11.phx.gbl...
> > > > After Microsoft's doing so well in the TPC raw performance bechmark
> for
> > so
> > > > long, Oracle appears to have finally leapfrogged SQL Server on both
> > > > Clustered and Non-Clustered configurations. In fact, not only has
> Oracle
> > > > leapfrogged them but literally blew SQL Server away.
> > > >
> > > > What is REALLY interesting about this is the fact that the
number-one
> > > > performance for the clustered configuration is both considerably
> higher
> > > than
> > > > Microsoft and WITH BETTER PRICE PERFORMANCE, which has always been
> > > > Microsoft's strength.
> > > >
> > > > This appears to be a function of using an Open Source OS (Linux).
> > > >
> > > > I still generally prefer SQL Server over Oracle (easier to tune and
> > > > maintain) but have been troubled by Microsoft's trend AWAY from
> > commodity
> > > > pricing (that they pioneered at the high end) at the same time
> > Enterprise
> > > > providers have been trending toward it. I'm wondering if this latest
> > > > development might single an inflection point where Microsoft begins
to
> > > > rethink ever-upward pricing and once again return to a more
aggresive
> > > > position in this area.
> > > >
> > > >
> > >
> > >
> >
> >
>|||Just look at the dates submitted and fact that Oracle has built (how many?) new versions since 1999 and releasing of MS SQL 2000!?sql