On Thu, Jul 15, 2010 at 1:51 PM, Tim Cook <t...@cook.ms> wrote: > Not to mention you've then got full-time staff on-hand to constantly be >replacing > parts.
Maybe I don't understand something, but we also had on-hand full-time staff to constantly replacing Dell's parts..., so what's the problem? Dell or HP or Sun are crashing exactly as same as SuperMicro machines (well, not really: Dell is more horrible, if you ask). Vendor, that sells us SuperMicro boxes offers as same support as we could get from HP or Dell. So all we do is simply pull out off the rack the thing and let vendor takes care of it. Machines are built automatically from the kickstart. What exactly I am missing then? > Your model doesn't scale for 99% of businesses out there. Unless > they're google, and they can leave a dead server in a rack for years, it's > an unsustainable plan. Not sure what you're talking about here, but if I run a cluster, then I am probably OK if some node[s] gone. :) Now, how it does not scales, if the vendor that works with IBM directly (in my case there is no real IBM in the über-country I am living but a third-party company that only merchandizing the name) came and took my hardware for repair. Vendor that works with the Dell (same situation) directly came and took my hardware for repair. Vendor that works with HP directly came and took my hardware for repair. Apple officially NOT repairing their XServe, but give parts to a third-party company that does the same to HP or IBM (!) or Dell or Supermicro — that happens in the country I am living, yes. And now the vendor that works directly with Supermicro took my hardware for repair on the same conditions as others. In any case, no matter what box (white, black, beige, silver, green, red, purple) I still experiencing: 1. A downtime of the box (obviously). 2. A chain of phonecalls to support, language of which could be more censored. 3. A vendor coming and taking a brick with himself. 4. A some time for repair taking a while. 5. A smile from the vendor, when they returning the box back to the DC. This sequence yields to all the vendors I've mentioned. Now, what exactly is the problem other than just scary grandma's stories that my model does not scales and big snow bear will eat me alive? I have to admit that I have no experience running 10K servers in one block like you do, so my respect is to you and I'd like to know the exact problems I might step into and the solution to avoid. Since you running this amount of machines, so you know it and you can share the experience. But from what I do have experience, I can not foresee some additional problems that we have with HP or Dell or Sun or IBM boxes. So could you please elaborate your statements? I would appreciate that (and some other folks here as well would be interested to listen to your lesson). Thank you. -- Kind regards, BM Things, that are stupid at the beginning, rarely ends up wisely. _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
