Well we were sucessful in getting past the first approvals to move a
"large" scale application from a variety of "platforms" into Rails.
Of course our database of choice is Oracle. The next step is
to "design' the total "infrasturure". We have decided we want to use
Oracle 11 on Linux. So later we can take advantage of "replication",
and maybe clustering support.
I've already got 8i working. Conversion scipts and migrations, so happy
with Oracle 10g in "first" prototype.
Now the next level is setting up the dedicated "database" server, dedicated
web server etc in as close as production as possible.
So what is oracle's stand on giving us a 11g "client" verses a version 8 client for rails?
The other "concept" is using a "shelf" of blades to scale up the database side.
Which I belive we are going to need a better client to support as well.
The change from 10g to 11g includes some features that will be very useful. The DRCP - database resident connection pools - is a big help in keeping the connection count down. There are also auto-tuning features on the db side that make managing the database easier. Managing a large-scale Oracle implementation usually requires a specialist, or a remote-db management contract with a for-hire specialist.
The OCI8 client naming isn't based on db version - both 10g and 11g use the OCI8 libraries and adapters. You can get the latest client from the OTN site.
The idea of using blades for the Oracle db is ok - assuming that the blade has the horsepower, memory and disk (or SAN/NAS connectivity) that it will need. The sticking point is bang for the buck - a maxed-out HP full-height Xeon or Opteron blade costs about the same as a Sun T5240 with the same amount of memory. The Sun gear is a much better performer for high db loads. And then there's the cost of the Oracle license - which is the same for both. Scaling blades out as Oracle engines is /very/ expensive compared to scaling up with a non-blade server.