Jim Kennedy wrote:
>"Domenic G." <domenicg_at_hotmail.com> wrote in message
>news:c7e08a19.0310051102.2ae576e8_at_posting.google.com...
>
>
>>Hi,
>>
>>I've been reading about 10g -- all this grid "hype" sounds like
>>Parallel Server (excuse me, RAC) rehashed with some BS tossed in to
>>boot.
>>
>>According to the grid concept, Oracle is supposed to be able to
>>connect maybe 10s to 100s?? of low-cost boxes together to form a
>>"grid" -- hmmm, does Parallel, I mean RAC go that high? Don't think
>>so, and I don't think that would be low cost!
>>
>>In Parallel Server, I mean RAC, you're connected to one instance --
>>does this mean that in 10g, a saturated parallel server instance
>>automatically redirects the request to another parallel server
>>instance in the cluster to harness "the power of the grid"? I didn't
>>know RAC could do this, maybe in 10g -- I know that when the instance
>>you're connected to goes down, if your TNS is set up right, you flip
>>over to the other instance and SELECTs will fail-over while write
>>transactions must be resubmitted. Concept also works with MM
>>replication except no transaction failover, only connections flip
>>over.
>>
>>So what is the true meat and potatoes of 10g GRID? -- is it a new
>>version of RAC that doesn't require hardware clusters? How can that
>>possibly work? I'm beginning to think this stuff is just empty hype.
>>Maybe the "g" should be for "Good Grief, Charlie Brown".
>>
>>Domenic.
>>
>>
>
>Grid is just the next step from RAC. It really does involve RAC but in more
>dimensions than you are thinking of it. Gee where to start. RAC in 9i was
>Oracle Parallel server , they did add an Oracle Managed File system which
>makes RAC easier. (which is also on the way to grid). Think of several
>layers. At one layer you have the "place where the data is physically
>stored", this can be NAS, SAN, or whatever other buzz word, but basically
>the data is on one shared place. Oracle Managed File System manages access
>to this for Oracle.(You don't have to go that way, but that is the intent).
>At the next layer you have the database. (Oracle RAC) The idea is that you
>can have a large selection of inexpensive boxes that access the high
>performance storage layer. If you need anther box, just bring it on line.
>(see one of the sessions at Oracle world where they had 16 nodes doing
>this - the idea was not to have an arms race but to see how easy or
>difficult it would be to setup and manage such a thing. Additionally, to
>see what failover, recovery, and how to load balance things. eg Data
>Warehouses and oltp which might have different needs in a 24 hour cycle -
>oltp peaks during the day and Data Warehouse peaks at night, could shift
>some machines from one to another instead of buying for peak load for each
>system.)
>
>The next level of the "grid" was to add application servers. (J2EE compliant
>apps) Oracle App Server would fit there. Then on top you have some sort of
>load balancer. You have OEM provision and shift loads for all of this stuff
>to manage the Grid. Sure is there some marketing hype in there, of course,
>one would be naive to not see that. But there are some good ideas in there
>also, and if they work - time will tell - they could be an advantage to a
>business.
>
>Jim
>
>
In addition there is still another aspect that should not be forgotten.
The data center of tomorrow will not
have the concept of this machine has a specific number of CPUs. IBM,
Sun, and HP are well on the way
to making that concept obsolete. So what is Oracle to do when users come
in from the web through a
finite number of connections into an instance with a dynamically
configurable numbers of CPUs? Licsense
by the number of parking spots in the lot? They had no choice but to go
to grid.
--
Daniel Morgan
http://www.outreach.washington.edu/ext/certificates/oad/oad_crs.asp
http://www.outreach.washington.edu/ext/certificates/aoa/aoa_crs.asp
damorgan_at_x.washington.edu
(replace 'x' with a 'u' to reply)
Received on Sun Oct 05 2003 - 15:30:17 CDT
