Informix on NUMA-Q machines

Does anyone run their OnLine database on a Sequent NUMA-Q machine? If so, can you suggest any online sources of information regarding performance comparisons between a standard SMP machine and a NUMA-Q machine? Or any online info that will explain how the NUMA architecture works from more than a marketing perspective? Specifically, I'm wondering about what happens when a query is broken up across two nodes and the results have to be shared across a Q-Link (?). Will this result in the same bottleneck as a standard SMP machine? Any guidance you can give in this regard will be greatly appreciated. Sincerely, Troy Peterson

Troy Peterson wrote: > > Does anyone run their OnLine database on a Sequent NUMA-Q machine? > > If so, can you suggest any online sources of information regarding > performance comparisons between a standard SMP machine and a NUMA-Q machine? > Or any online info that will explain how the NUMA architecture works from > more than a marketing perspective? > > Specifically, I'm wondering about what happens when a query is broken up > across two nodes and the results have to be shared across a Q-Link (?). > Will this result in the same bottleneck as a standard SMP machine? > > Any guidance you can give in this regard will be greatly appreciated. There is some information on tuning and configuring for a Numa architecture machine, based on my experiences with DG Numa systems, in my article for Informix Tech Notes (V8 No3). Art S. Kagel