I don't know that it's a "best practice", but each volume gets its own affinity, and you can have parallelization issues. That said, I've seen an A700 on FCP with multiple LUNs hitting 7 GB/s on all reads. We have some optimizations starting in 9.3 which make this better.
The biggest thing is to have multiple LUNs I'd say either way per host for performance. On a 2750 this is probably not a big deal since it's a low end platform, but if you can stripe your LUNs you'll get best performance from a threading standpoint.