Today, more reader input, this time on my newsletter regarding multi-core systems. Several readers wrote to tell me that, yes, the move to multi-core platforms obviously is inevitable, but there remain several wrinkles. Those include issues about reliability, security and I/O throughput with more workloads being put onto single boxes.

Al Muller, a principal architect at AT&T Consulting and the author of a pair of books on virtualization, says that multi-core technology will be the foundation for consolidating proliferating hardware into a more energy- and cost-efficient virtualized environment.

“As more processing power becomes available in 1U, 2U and 4U form factors, the ability to maximize consolidation ratios onto these traditional servers will dramatically increase,” he says.

He notes that he just completed a virtualization design for an international publishing company and by using dual-core AMD processors, reduced their hardware needs by 25%. “Had true multi-core processors been available, the ratio could have been driven even higher, increasing efficiency, decreasing power and cooling requirements and reducing hardware and data center space costs.”

Along with advances in the server processor, however, there needs to be other hardware improvements in order to gain the greatest benefit out of the multi-core systems, Muller says.

“Other hardware advances are important, such as higher RAM concentrations per server, faster network connectivity, dual- and quad-port NICs and faster access to storage such as 4GB HBAs,” he says.

Another reader says he also is looking at multi-core systems, but taking things slowly.

“For non-mission critical work it's a no-brainer ….move it to multi-core systems,” says Bob Williams. “For complex systems that are mission critical, co-hosting them on the same physical unit is being approached with a great deal of caution.”

He adds: “What disturbs me most is that we are now starting to see the real need to solve mainframe like problems on small/mid-sized systems doing mainframe work. It is, frankly, scary. For example, mainframe system OSes and [mission critical] applications have a feature called checkpoint re-start. Co-hosting a number of [mission critical] applications without this capability [or something similar] is playing roulette with a guaranteed outcome. You are going to get nailed.”