> No. BTDT. It depends on requirements for HW quality. Intel penguin farm > could mean high-end xSeries or blades, or simply bunch of regular PCs > (see Google).
Actually, it's not the hardware costs that make the TCO comparison favorable to System z Linux. It's typically the savings in software licenses and people productivity. The bottom line, however, is if you only have a relatively small number of systems, running software that is either inexpensive, and/or not priced by the processor, have plenty of power, cooling and floor space, moving to Linux on System z won't make any sense. > I glanced at Sine Nomine's page about OpenSolaris for Z. There's a > prominent restriction that it runs only under z/VM, not in an LPAR. > So it exploits a CP feature. An easy conjecture, with no evidence, > is that it uses CP Block DASD I/O to bypass the complexities of > CKD channel programs. It uses a lot of features for various reasons, not just the DASD code. The design goal from the beginning was to maximize the amount of interaction between OpenSolaris and z/VM. > Then, is it fairer to compare VMWare to z/VM or to PR/SM? It would still compare very unfavorably. > Is OpenSolaris for z eligible for IFL? Yes, although "eligible" is misleading, given the licensing of it. OpenSolaris will run on an IFL or a standard CP. > From what little I understand, any OS which does not invoke some specific > functions (undocumented) should work on an IFL. It's actually a single instruction, not even a diagnose. > I was just trying to point out that hardware/environmental cost > comparisons do not present the whole picture. > There have been several presentations at SHARE and EXPO from companies > realizing significant savings implementing Linux on z. The major savings are almost always in software licensing costs and people productivity. > The big thing that seems to be missing is a comparison of virtualized > Intel vs. IFL/z/VM. There is no comparison. When I tell people that we have customers running multiple hundreds of Linux instances on single boxes with many CPUs, they almost go into shock. > I thought the z10 4.4GHz chips were the answer to the CPU intensive > issue (but probably not all). Yes and no. The CPUs are faster, but they're also orders of magnitude more expensive. So, while you might be able to get better performance from a z10, the cost per computation is still going to be much higher. Mixed workloads are still the best fit. > The larger question is can IFL/z/VM compete with Intel/Vmware (or > equivalent) when ALL costs are considered. That's the only way Linux for System z can compete. If you leave out a few cost factors, System z loses. It's the fact that most people _don't_ look at all the cost factors that distributed systems look good when they start getting rather numerous. > I think they are dreaming. Sure, they can support Fibre Channel, but how > many of them? Certainly not hundreds. The memory on a PC can't handle that > kind of data rate. It's not the memory that's the limiting factor. PC memory is extremely fast. It's the CPU combined with the (lack of) I/O subsystem. The same problem exists on SPARC, PPC, etc. > That is interesting. But I am wondering about so-called Oracle Site > Licenses. From what I have been told, we have such a thing. Supposedly, this > allows us to have basically unlimited numbers of Oracle systems running on > any number of "cores" on any platform. That's how we got Oracle on z/OS. It > was "free" with out current license. But I've been misinformed before. Those are normally called Enterprise License Agreements, and the price you pay varies by the number of licenses you buy, and they get "trued up" periodically. The reason why people pay for ELAs is because the cost of buying a huge amount of individual licenses is killing them. If the number of licenses needed/anticipated drops by an order of magnitude, you wouldn't need to pay for an ELA in the first place, just the smaller number of individual licenses. > But, you've got power and cooling of one zFOOTPRINT vs all the other INTEL > boxen. While this number can be large in absolute terms, it's a tiny percentage of the TCO calculation. (At least until you run out of them or floorspace. Then you're looking at the $25M server.) ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
