Maximum length of aound 750 bp could be achieved with 454 GS-FLX Titanium platform. Other sequencing platforms give shorter read lengths around 1500 bp. In between the conversed region, this 16s RNA gene has about 9 hypervariable regions. 454 pyroseq. platform is most common for amplicon based sequencing. Here the target specific primers are linked to std. 19mer adapter sequencing primers. It anneals the amplicons to capture beads. Various adapter primers r linked to target primers which get amplified in opoosite direction. Later to which multiplex identifier could be added up.

Classically, 2 targets are chosen concerning hypervariable regions in pyrosequencing:

V1-V3 regions, or V3-V5 regions. You can have a look on the protocol given by the HMP consortium, that made quite a job for normalizing things! (http://www.hmpdacc.org/doc/16S_Sequencing_SOP_4.2.2.pdf).

Also, it is important to have a look at what was used before in the context of your study. By this, I mean that it is always interesting to compare your results with previous studies, but it is feasable only if you target the same 16S regions.

Finally, I think that you can take a decision by looking at the universality of the different primers available. Take extra care of that if you want to have Archae.

Hope it'll help!

Good luck for your work!

Hope this paper can help. And seeking for longer reads from 454 or Iontorrent may not be a good choice as the sequencing quality goes all the way down after 300bp, which will result in large number of sequences being filtered out. 

Figure 1 in http://www.ncbi.nlm.nih.gov/pubmed/20804791 will give you the information you need - relative variability of the V regions in the SSU of bacteria, archaea, and eukarya.