Take that chart we've been looking at for years: on one axis you have cost; on the other you have topic X. X is usually restore time, or redundancy, or reliability, and so on. The idea is If you want more of X, you've have to be willing to spend
...more>
Take that chart we've been looking at for years: on one axis you have cost; on the other you have topic X. X is usually restore time, or redundancy, or reliability, and so on. The idea is If you want more of X, you've have to be willing to spend more. Let's make X reliability. What happens when the reliability side of that chart flattens? Where cost doesn't necessarily equate to significantly more reliability? Then you pick the solution that costs less with the same results.
Recently, AIX/Power systems achieved top spot in the ITIC survey referenced in this article for reliability. It highlights that participants reported only 15 minutes of unplanned downtime per year for those systems. HOWEVER, check out this quote: “The second most reliable servers in terms of downtime were customized versions of Novell SuSE Linux running on standard x86 hardware, clocking in at 17.4 minutes of downtime per year.”
![]()
Granted, it states 'customized' version of SLES, which I'd like to understand more, but really - 2.4 minutes of downtime difference over an entire year for second place? Another way to put it: 99.99715% vs 99.99669% reliability. (15 / minutes per year and 17.4 / minutes per year)
It would be interesting to understand the cost comparison of the top two systems in the survey. I think it's safe to say that it is more expensive to implement the POWER architecture solutions over the Linux x86 solutions from a hardware perspective (yeah, I know, I just threw out some serious flamebait for you RISC fans). It is what it is: typically RISC is more expensive than x86 hardware. The question is how much more expensive is it for those two minutes of difference in reliability over an entire year?
There are, of course, other factors for an understanding of true 'total cost of ownership' (TCO) including software, staffing, and more. In years past TCO, especially as it relates to software licensing, was a valid argument for the POWER systems line. Fewer processors to get the same job done = less software core licensing costs. However, taking into account performance of recent Intel/AMD cores and adjusted software licensing value units for specific core architecture, that argument is a thing of the past.
Of course, if you went with the AIX solution, you'd have 2.4 more minutes of time to do things like read my blog post; yet your budget for something else might suffer --i.e. travel to VMworld to meet with all of us on the DellTechCenter team at our booth.
less>