Since the issue came up I checked out some references on the performance of oracle (10g) on t1000/t2000. Here is some wisdom found via google.
Basic oracle stuff:
* oltp scales well on niagara - no floating point, typically i/o bound
* a bit of tuning to exploit maximal parallelism may be necessary
* lots of parallel queries change the profile from i/o bound to cpu bound
* an obscure oracle jvm trap could hang the installation procedure
* don't copy/clone an existing sparc deployment onto a t1/t2
* do you want oracle apps running on the same h/w?
* the t2k does well in the space/watts/performance arena