Quantitative RAS Comparison of Sun CMT/Solaris and X86/Linux Servers

By incorporating the Chip Multi-Threading (CMT) and operating system predictive self-healing technologies, the Sun CMT/Solaris based servers are not only cost/performance effective, but also more robust in reliability, availability, and serviceability (RAS) than the X86/Linux servers with similar performance. The differentiators include higher levels of hardware integration, more fault tolerance provisions in processors, the Solaris memory page retirement (MPR), and the Solaris/SPARC processor offlining (PO) capabilities of the CMT/Solaris server. This study applies analytical models, with parameters calibrated by field experience, to quantitatively compare system RAS, against hardware faults, between the CMT/Solaris and X86/Linux servers. The results show significant RAS benefits of the CMT, MPR, and PO technologies.