My recent clients have been relying on VMs (VCE/Vblock, AIX/lpar) to manage cpu resources, to include running ASE and RS on their own VMs.
While this means a specific ASE/RS pair may not be co-located on the same 'machine', in reality they are usually running on the same frame (or next to each other in the same computer center) and the network connectivity between the VMs is pretty good (<1ms) ... so co-locating the ASE and RS on the same 'machine' hasn't been a major concern ... ymmv.
Are your RH hosts running on physical or virtual machines?
----------------
My current client is running RS 15.7.1 on AIX/lpars.
The lpars have 2 or 3 'cores' (as seen by RS/SYSAM licensing).
We're running upwards of 5 RSs per lpar, with each RS managing 35-75 ASE connections.
All RS instances are running with smp_enable=on. We do see high cpu utilization during EOD batch (and release) processing, so the high cpu usage is somewhat expected.