Can SQL Server be load balanced ?

Can one use any h/w load balancing to balance a farm of SQL Servers ? I am
trying to figure if its possible maybe on some read only servers...or maybe
even on OLTP..Can someone help me here with some articles or thoughts on
this ?So is this basically for read only SQL servers ? If also for OLTP, how would
you load balance between 2 SQL Servers that have transactions maybe
committed on one server and not on the other ? Is there an article for SQL
Load balancing ?
"Denny" <mrdenny@.gamespy.com> wrote in message
news:%23H46kTKgDHA.2072@.TK2MSFTNGP10.phx.gbl...
> Yes. I've used the Windows Load Balancing service to ballance 2 SQL
Server.
> I don't see why you couldn't use a h/w load balancer as well.
> --
> Denny Cherry
> DBA
> GameSpy Industries
> "Hassan" <fatima_ja@.hotmail.com> wrote in message
> news:uQ%23ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > Can one use any h/w load balancing to balance a farm of SQL Servers ? I
am
> > trying to figure if its possible maybe on some read only servers...or
> maybe
> > even on OLTP..Can someone help me here with some articles or thoughts on
> > this ?
> >
> >
>|||You cannot load balance an OLTP database (not in the sense of dynamic load
balancing which redirects a user request to whichever server that has a
'ligher' load at the moment)
--
Linchi Shea
linchi_shea@.NOSPAMml.com
"Hassan" <fatima_ja@.hotmail.com> wrote in message
news:uCLYgcKgDHA.2384@.tk2msftngp13.phx.gbl...
> So is this basically for read only SQL servers ? If also for OLTP, how
would
> you load balance between 2 SQL Servers that have transactions maybe
> committed on one server and not on the other ? Is there an article for SQL
> Load balancing ?
> "Denny" <mrdenny@.gamespy.com> wrote in message
> news:%23H46kTKgDHA.2072@.TK2MSFTNGP10.phx.gbl...
> > Yes. I've used the Windows Load Balancing service to ballance 2 SQL
> Server.
> > I don't see why you couldn't use a h/w load balancer as well.
> >
> > --
> > Denny Cherry
> > DBA
> > GameSpy Industries
> > "Hassan" <fatima_ja@.hotmail.com> wrote in message
> > news:uQ%23ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > > Can one use any h/w load balancing to balance a farm of SQL Servers ?
I
> am
> > > trying to figure if its possible maybe on some read only servers...or
> > maybe
> > > even on OLTP..Can someone help me here with some articles or thoughts
on
> > > this ?
> > >
> > >
> >
> >
>|||Hi Hassan.
SQL Server definitely CAN be load balanced under Windows Load Balanced
Clustering for READ ONLY. This can be accomplished by installing identical
(or even similar) installations of the same database on many machines,
perhaps with replication or log shipping between them to collect updates
from external sources.
You might even configure SQL Servers for pure data collection under a
Windows Load Balancing Cluster as well.
You can NOT clustered SQL Server for load balancing for OLTP applications
though, as this requires a distributed lock manager & interconnect
technology on top of shared disks design that SQL Server doesn't have. It's
a different clustering architecture that some other RDBMS vendors do
implement, & certainly far more complex than SQL Server's shared nothing
(Federated) approach.
Is this a good / bad thing? This argument is often taken up with vigour by
other platform afficionados but imho, no clustering approach wins on all
fronts. Federated clustering (SQL Server's approach) has proved scalability
(TPC results) but Oracle's top clutered benchmark results are (far) lower
than their unclustered results! Perhaps this just points to industry appathy
toward clustered solutions in favour of single box solutions, mitigating the
whole argument...
From my perspective, I'm seeing much more server consolidation than
clustering going on in the marketplace. This might just be a local trend
here in Australia, but I've rarely seen clustering implemented, either on
Oracle or SQL Server other than for redundancy / availability.
Regards,
Greg Linwood
SQL Server MVP
"Hassan" <fatima_ja@.hotmail.com> wrote in message
news:uQ#ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> Can one use any h/w load balancing to balance a farm of SQL Servers ? I am
> trying to figure if its possible maybe on some read only servers...or
maybe
> even on OLTP..Can someone help me here with some articles or thoughts on
> this ?
>|||dont forget that a form of load balancing can also be achieved by using
horizontal partitions (which can be R/W)
--
BR,
Mark Broadbent mcdba,mcse+i
_________________________
"Greg Linwood" <g_linwood@.hotmail.com> wrote in message
news:%23PryWMPgDHA.3208@.TK2MSFTNGP11.phx.gbl...
> Hi Hassan.
> SQL Server definitely CAN be load balanced under Windows Load Balanced
> Clustering for READ ONLY. This can be accomplished by installing identical
> (or even similar) installations of the same database on many machines,
> perhaps with replication or log shipping between them to collect updates
> from external sources.
> You might even configure SQL Servers for pure data collection under a
> Windows Load Balancing Cluster as well.
> You can NOT clustered SQL Server for load balancing for OLTP applications
> though, as this requires a distributed lock manager & interconnect
> technology on top of shared disks design that SQL Server doesn't have.
It's
> a different clustering architecture that some other RDBMS vendors do
> implement, & certainly far more complex than SQL Server's shared nothing
> (Federated) approach.
> Is this a good / bad thing? This argument is often taken up with vigour by
> other platform afficionados but imho, no clustering approach wins on all
> fronts. Federated clustering (SQL Server's approach) has proved
scalability
> (TPC results) but Oracle's top clutered benchmark results are (far) lower
> than their unclustered results! Perhaps this just points to industry
appathy
> toward clustered solutions in favour of single box solutions, mitigating
the
> whole argument...
> From my perspective, I'm seeing much more server consolidation than
> clustering going on in the marketplace. This might just be a local trend
> here in Australia, but I've rarely seen clustering implemented, either on
> Oracle or SQL Server other than for redundancy / availability.
> Regards,
> Greg Linwood
> SQL Server MVP
> "Hassan" <fatima_ja@.hotmail.com> wrote in message
> news:uQ#ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > Can one use any h/w load balancing to balance a farm of SQL Servers ? I
am
> > trying to figure if its possible maybe on some read only servers...or
> maybe
> > even on OLTP..Can someone help me here with some articles or thoughts on
> > this ?
> >
> >
>|||"Greg Linwood" <g_linwood@.hotmail.com> wrote in message news:<#PryWMPgDHA.3208@.TK2MSFTNGP11.phx.gbl>...
> You can NOT clustered SQL Server for load balancing for OLTP applications
> though, as this requires a distributed lock manager & interconnect
> technology on top of shared disks design that SQL Server doesn't have. It's
> a different clustering architecture that some other RDBMS vendors do
> implement, & certainly far more complex than SQL Server's shared nothing
> (Federated) approach.
DB/2 UDB is also shared nothing & can be clustered for OLAP or OLTP.
Shared-nothing DB/2 UDB avoids the interconnect overhead seen in
Oracle clusters. The latest version of DB/2 enables this with relative
ease and scales OLTP installations reasonably well. Using a "shared"
SAN appliance (separate RAID box) solves the issue of shared-nothing
fail-over.
Why MS hasn't been able to actually cluster MS SQL is a bit of a
mystery to me, perhaps it goes back to legacy issues that still limit
Sybase as well.
> "Hassan" <fatima_ja@.hotmail.com> wrote in message
> news:uQ#ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > Can one use any h/w load balancing to balance a farm of SQL Servers ? I am
> > trying to figure if its possible maybe on some read only servers...or
> > maybe even on OLTP..Can someone help me here with some articles or
> > thoughts on this ?
I've never found it easy to get a clear concise document from MS that
stated exactly what kind of clustering SQL Server was capable of...
you have to dig & ask around... ;-)
Interestingly, MySQL has InnoDB that enables OLTP clustering (I think
it's like Oracle...), and MySQL is merging its codebase with SAP-DB
(both open source). SAP opened sourced SAP-DB a couple of years ago,
it is a lighter version of Adabas but has many of the transactional
features currently missing from MySQL (tran logs, triggers, column
constraints, ref. integrity, stored procedures...). Methinks MS had
better get their act together & come up with a full shared-nothing
OLTP capability.
/lee|||One option is Distributed Partitioned Views, see the resource kit chapter at
http://www.microsoft.com/technet/prodtechnol/sql/reskit/sql2000/part10/c3861.asp
--
Hal Berenson, SQL Server MVP
True Mountain Group LLC
"Hassan" <fatima_ja@.hotmail.com> wrote in message
news:uCLYgcKgDHA.2384@.tk2msftngp13.phx.gbl...
> So is this basically for read only SQL servers ? If also for OLTP, how
would
> you load balance between 2 SQL Servers that have transactions maybe
> committed on one server and not on the other ? Is there an article for SQL
> Load balancing ?
> "Denny" <mrdenny@.gamespy.com> wrote in message
> news:%23H46kTKgDHA.2072@.TK2MSFTNGP10.phx.gbl...
> > Yes. I've used the Windows Load Balancing service to ballance 2 SQL
> Server.
> > I don't see why you couldn't use a h/w load balancer as well.
> >
> > --
> > Denny Cherry
> > DBA
> > GameSpy Industries
> > "Hassan" <fatima_ja@.hotmail.com> wrote in message
> > news:uQ%23ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > > Can one use any h/w load balancing to balance a farm of SQL Servers ?
I
> am
> > > trying to figure if its possible maybe on some read only servers...or
> > maybe
> > > even on OLTP..Can someone help me here with some articles or thoughts
on
> > > this ?
> > >
> > >
> >
> >
>|||Just priorities and time. It's interesting to think that MS could easily
add clustering, but you have to put the existing implementations in
perspective. Both IBM and Oracle started their clustering work in the 80s,
and only recently have they gotten them to be useful. Microsoft didn't even
start SQL Server development until the mid-90s and has had lots more
important work to do before considering clusters. So good clustering is
hard to do, and you have to make it a priority before it happens.
IBM, for example, was producing a large loosely-coupled multiprocessor
called the SP2 and needed software in order to be able to sell it. So they
commissioned a sizable team to build a totally custom version of DB2 for the
SP2 called DB2 Parallel Edition. Then IBM had a project to integrate a
bunch of the different DB2 code bases together, thus bringing the PE changes
into the mainstream code base and giving them a general clustering
capability. To date it has proven very successful at Data Warehousing (the
original SP2 target), but not at OLTP. It can do OLTP, but that's not its
forte. I don't know of anyone running their ERP system on a DB2 cluster,
for example.
Oracle needed a competitive response to DEC Rdb's clustering capability and
produced a poor clone of that capability, OPS. The purpose of DEC and
(Oracle's) implementation was to allow scalability beyond the largest DEC
VAX system available. The availability characteristics were even more
important. Even DEC's superior implementation really only scaled to 3 or
maybe 4 systems for OLTP, but 4 mainframe-class systems clustered together
was more than any application needed. Oracle has re-purposed OPS several
times in its life, with very limited success until just a few years ago.
Two things happened along the way. First, Oracle realized they could use
OPS (now renamed RAC) for fast failover. Most OPS/RAC customers actually
use it just for this and not for scalability. Second, Oracle has recently
improved OPS/RAC so that it scales (and is manageable) for a growing number
of applications. I haven't looked at 10G, but I would expect there has been
further progress in this area.
The key for both IBM and Oracle's clustering capabilities is that they are
primarily aimed at the high end. Their use is generally only indicated when
you outgrow the largest single system box you can purchase. This may be
changing, but it is the historical situation.
All the questions about "load balancing" in this newsgroup have nothing to
do with the IBM or Oracle-type targets. They aren't about "My 32-processor
Unisys ES/7000 or 64-processor HP Superdome is out of steam, what should I
do?" They aren't even about running out of steam on an 8-processor box,
which is the largest "commodity" system. The questions here are about how
do I load balance two or more dual-processor x86 systems? That's not
something anyone has really targetted. You could do it with either IBM or
Oracle technology, but that isn't their design center and you will
undoubtedly be dissapointed. Total cost will be higher, operational costs
will be much higher, and total performance much lower, than just buying a
bigger SMP machine.
As for Microsoft, they have just had higher priorities then doing full-blown
clustering. And with the collapse of the internet bubble much of the market
demand also went away. In 1999-2000 the market was saying "My business is
growing so fast that even a single Sun UE10000 (the biggest server of the
day) can't handle the growth, so I need the option to cluster them". That
was the justification for Oracle's push to turn OPS into RAC. Microsoft was
certainly struggling with 8-way systems being the largest available for
running Windows, and that's where Distributed Partitioned Views came from.
But soon after the Unisys ES/7000 arrived, giving Microsoft a high-end
32-way x86 box to run on, and the Internet bubble burst bringing reality
into company's growth projections. Now there is the HP Integrity Superdome
system with 64 Itanium 2 processors. This is a box that allows SQL Server
to handle almost 800,000 transactions per minute at a lower cost per
transaction then any cluster has demonstrated. How many applications do
you know of that need more than 800,000 transactions per minute? Where is
the market pressure for scalable clusters?
I think that there is a very interesting market for clustering low-end
servers, and it's one that Microsoft should address. They already do this
in special cases, for example Replication let's you handle numerous
scale-out scenarios. But at some point it would be nice if they allowed you
to say "here are these 4 server boxes, make them act like one large server"
and have it work automagically for a large number of applications. This
could be made to work well for 4 dual-processor machines, or 4 Superdomes.
I'm a cluster fan, so my reading of the tea leaves is that Microsoft should
do this sooner rather than later. I don't know what the powers that be at
Microsoft are seeing in the tea leaves.
--
Hal Berenson, SQL Server MVP
True Mountain Group LLC
"Leebert" <buddhashortfatguy@.yahoo.com> wrote in message
news:e2d95f55.0309220741.221a6314@.posting.google.com...
> "Greg Linwood" <g_linwood@.hotmail.com> wrote in message
news:<#PryWMPgDHA.3208@.TK2MSFTNGP11.phx.gbl>...
> > You can NOT clustered SQL Server for load balancing for OLTP
applications
> > though, as this requires a distributed lock manager & interconnect
> > technology on top of shared disks design that SQL Server doesn't have.
It's
> > a different clustering architecture that some other RDBMS vendors do
> > implement, & certainly far more complex than SQL Server's shared nothing
> > (Federated) approach.
> DB/2 UDB is also shared nothing & can be clustered for OLAP or OLTP.
> Shared-nothing DB/2 UDB avoids the interconnect overhead seen in
> Oracle clusters. The latest version of DB/2 enables this with relative
> ease and scales OLTP installations reasonably well. Using a "shared"
> SAN appliance (separate RAID box) solves the issue of shared-nothing
> fail-over.
> Why MS hasn't been able to actually cluster MS SQL is a bit of a
> mystery to me, perhaps it goes back to legacy issues that still limit
> Sybase as well.
> > "Hassan" <fatima_ja@.hotmail.com> wrote in message
> > news:uQ#ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > > Can one use any h/w load balancing to balance a farm of SQL Servers ?
I am
> > > trying to figure if its possible maybe on some read only servers...or
> > > maybe even on OLTP..Can someone help me here with some articles or
> > > thoughts on this ?
> I've never found it easy to get a clear concise document from MS that
> stated exactly what kind of clustering SQL Server was capable of...
> you have to dig & ask around... ;-)
> Interestingly, MySQL has InnoDB that enables OLTP clustering (I think
> it's like Oracle...), and MySQL is merging its codebase with SAP-DB
> (both open source). SAP opened sourced SAP-DB a couple of years ago,
> it is a lighter version of Adabas but has many of the transactional
> features currently missing from MySQL (tran logs, triggers, column
> constraints, ref. integrity, stored procedures...). Methinks MS had
> better get their act together & come up with a full shared-nothing
> OLTP capability.
> /lee|||"Hal Berenson" <haroldb@.truemountainconsulting.com> wrote in message news:<egL85WfgDHA.1684@.TK2MSFTNGP10.phx.gbl>...
> The questions here are about how
> do I load balance two or more dual-processor x86 systems? That's not
> something anyone has really targetted. You could do it with either IBM or
> Oracle technology, but that isn't their design center and you will
> undoubtedly be dissapointed.
But isn't that what Oracle was pushing with RAC, x86 blades? I think
Oracle's performance would be poorer than DB2's (which we found to be
good), but the admin more dynamic and flexible than DB2.
> Total cost will be higher, operational costs
> will be much higher, and total performance much lower, than just buying a
> bigger SMP machine.
Ahhhh, Moore's curve.
Two years ago we saw decent scaling OLTP performance w/ DB2 UDB on
Linux... but 2 years ago PCs could only have 4 - 8 processors. The DB2
UDB partitioning key options were limited at the time, but to load
balance best we decided we wanted a relatively random distribution of
the data across the partitions, so we tuned the data extents *small*!
There are two goals really, one is scaling out, the other is HA (high
availability).
If you can get HA on the hardware level instead of software, so much
the better, but the idea of stacking cheap blade-servers (pushed in
Oracle RAC) is very appealing. RAC is more appealing for blade-servers
b/c it is so flexible (starting up new instances on-demand), DB2 UDB
isn't as flexible even yet.
Also I'm wondering... w/ gig-eth or net fabric, are Oracle's
interconnect mgm't overheads tolerable?
> This is a box that allows SQL Server
> to handle almost 800,000 transactions per minute at a lower cost per
> transaction then any cluster has demonstrated. How many applications do
> you know of that need more than 800,000 transactions per minute? Where is
> the market pressure for scalable clusters?
!!!!!
But that's a mega-$ machine. Sounds like a budget-buster.
I think many IT shops would like more in the way of scalable HA
clusters, but have it OK to live without them. Fortunately PCs are
reliable enough and the relative downtime and MTBF tolerable.
> I think that there is a very interesting market for clustering low-end
> servers, and it's one that Microsoft should address.
This is what Oracle's pushing with RAC... chain-gang lots of low-end
blades.
/lee|||Hi Hal..
<snip>
Oracle needed a competitive response to DEC Rdb's clustering capability and
produced a poor clone of that capability, OPS. The purpose of DEC and
(Oracle's) implementation was to allow scalability beyond the largest DEC
VAX system available.
</snip>
When I was at DEC, I unfortunately never really developed much of an
understanding of Rdb clustering technology, but I thought that it was
heavily hooked into VMS clustering (SSI). Is that correct? & if so, would
the limitation of current RDBMS clustering technology be tied to limits in
the current O/S clustering technolgies?
Regards,
Greg Linwood
SQL Server MVP
"Hal Berenson" <haroldb@.truemountainconsulting.com> wrote in message
news:egL85WfgDHA.1684@.TK2MSFTNGP10.phx.gbl...
> Just priorities and time. It's interesting to think that MS could easily
> add clustering, but you have to put the existing implementations in
> perspective. Both IBM and Oracle started their clustering work in the
80s,
> and only recently have they gotten them to be useful. Microsoft didn't
even
> start SQL Server development until the mid-90s and has had lots more
> important work to do before considering clusters. So good clustering is
> hard to do, and you have to make it a priority before it happens.
> IBM, for example, was producing a large loosely-coupled multiprocessor
> called the SP2 and needed software in order to be able to sell it. So
they
> commissioned a sizable team to build a totally custom version of DB2 for
the
> SP2 called DB2 Parallel Edition. Then IBM had a project to integrate a
> bunch of the different DB2 code bases together, thus bringing the PE
changes
> into the mainstream code base and giving them a general clustering
> capability. To date it has proven very successful at Data Warehousing
(the
> original SP2 target), but not at OLTP. It can do OLTP, but that's not its
> forte. I don't know of anyone running their ERP system on a DB2 cluster,
> for example.
> Oracle needed a competitive response to DEC Rdb's clustering capability
and
> produced a poor clone of that capability, OPS. The purpose of DEC and
> (Oracle's) implementation was to allow scalability beyond the largest DEC
> VAX system available. The availability characteristics were even more
> important. Even DEC's superior implementation really only scaled to 3 or
> maybe 4 systems for OLTP, but 4 mainframe-class systems clustered together
> was more than any application needed. Oracle has re-purposed OPS several
> times in its life, with very limited success until just a few years ago.
> Two things happened along the way. First, Oracle realized they could use
> OPS (now renamed RAC) for fast failover. Most OPS/RAC customers actually
> use it just for this and not for scalability. Second, Oracle has recently
> improved OPS/RAC so that it scales (and is manageable) for a growing
number
> of applications. I haven't looked at 10G, but I would expect there has
been
> further progress in this area.
> The key for both IBM and Oracle's clustering capabilities is that they are
> primarily aimed at the high end. Their use is generally only indicated
when
> you outgrow the largest single system box you can purchase. This may be
> changing, but it is the historical situation.
> All the questions about "load balancing" in this newsgroup have nothing to
> do with the IBM or Oracle-type targets. They aren't about "My
32-processor
> Unisys ES/7000 or 64-processor HP Superdome is out of steam, what should I
> do?" They aren't even about running out of steam on an 8-processor box,
> which is the largest "commodity" system. The questions here are about how
> do I load balance two or more dual-processor x86 systems? That's not
> something anyone has really targetted. You could do it with either IBM or
> Oracle technology, but that isn't their design center and you will
> undoubtedly be dissapointed. Total cost will be higher, operational costs
> will be much higher, and total performance much lower, than just buying a
> bigger SMP machine.
> As for Microsoft, they have just had higher priorities then doing
full-blown
> clustering. And with the collapse of the internet bubble much of the
market
> demand also went away. In 1999-2000 the market was saying "My business is
> growing so fast that even a single Sun UE10000 (the biggest server of the
> day) can't handle the growth, so I need the option to cluster them". That
> was the justification for Oracle's push to turn OPS into RAC. Microsoft
was
> certainly struggling with 8-way systems being the largest available for
> running Windows, and that's where Distributed Partitioned Views came from.
> But soon after the Unisys ES/7000 arrived, giving Microsoft a high-end
> 32-way x86 box to run on, and the Internet bubble burst bringing reality
> into company's growth projections. Now there is the HP Integrity
Superdome
> system with 64 Itanium 2 processors. This is a box that allows SQL Server
> to handle almost 800,000 transactions per minute at a lower cost per
> transaction then any cluster has demonstrated. How many applications do
> you know of that need more than 800,000 transactions per minute? Where is
> the market pressure for scalable clusters?
> I think that there is a very interesting market for clustering low-end
> servers, and it's one that Microsoft should address. They already do this
> in special cases, for example Replication let's you handle numerous
> scale-out scenarios. But at some point it would be nice if they allowed
you
> to say "here are these 4 server boxes, make them act like one large
server"
> and have it work automagically for a large number of applications. This
> could be made to work well for 4 dual-processor machines, or 4 Superdomes.
> I'm a cluster fan, so my reading of the tea leaves is that Microsoft
should
> do this sooner rather than later. I don't know what the powers that be at
> Microsoft are seeing in the tea leaves.
> --
> Hal Berenson, SQL Server MVP
> True Mountain Group LLC
>
> "Leebert" <buddhashortfatguy@.yahoo.com> wrote in message
> news:e2d95f55.0309220741.221a6314@.posting.google.com...
> > "Greg Linwood" <g_linwood@.hotmail.com> wrote in message
> news:<#PryWMPgDHA.3208@.TK2MSFTNGP11.phx.gbl>...
> >
> > > You can NOT clustered SQL Server for load balancing for OLTP
> applications
> > > though, as this requires a distributed lock manager & interconnect
> > > technology on top of shared disks design that SQL Server doesn't have.
> It's
> > > a different clustering architecture that some other RDBMS vendors do
> > > implement, & certainly far more complex than SQL Server's shared
nothing
> > > (Federated) approach.
> >
> > DB/2 UDB is also shared nothing & can be clustered for OLAP or OLTP.
> > Shared-nothing DB/2 UDB avoids the interconnect overhead seen in
> > Oracle clusters. The latest version of DB/2 enables this with relative
> > ease and scales OLTP installations reasonably well. Using a "shared"
> > SAN appliance (separate RAID box) solves the issue of shared-nothing
> > fail-over.
> >
> > Why MS hasn't been able to actually cluster MS SQL is a bit of a
> > mystery to me, perhaps it goes back to legacy issues that still limit
> > Sybase as well.
> >
> > > "Hassan" <fatima_ja@.hotmail.com> wrote in message
> > > news:uQ#ICOIgDHA.1700@.TK2MSFTNGP10.phx.gbl...
> > > > Can one use any h/w load balancing to balance a farm of SQL Servers
?
> I am
> > > > trying to figure if its possible maybe on some read only
servers...or
> > > > maybe even on OLTP..Can someone help me here with some articles or
> > > > thoughts on this ?
> >
> > I've never found it easy to get a clear concise document from MS that
> > stated exactly what kind of clustering SQL Server was capable of...
> > you have to dig & ask around... ;-)
> >
> > Interestingly, MySQL has InnoDB that enables OLTP clustering (I think
> > it's like Oracle...), and MySQL is merging its codebase with SAP-DB
> > (both open source). SAP opened sourced SAP-DB a couple of years ago,
> > it is a lighter version of Adabas but has many of the transactional
> > features currently missing from MySQL (tran logs, triggers, column
> > constraints, ref. integrity, stored procedures...). Methinks MS had
> > better get their act together & come up with a full shared-nothing
> > OLTP capability.
> >
> > /lee
>