Forums before death by AOL, social media and spammers... "We can't have nice things"
|    comp.os.vms    |    DEC's VAX* line of computers & VMS.    |    264,096 messages    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
|    Message 263,735 of 264,096    |
|    =?UTF-8?Q?Arne_Vajh=C3=B8j?= to All    |
|    Re: And so? (VMS/XDE)    |
|    12 Nov 25 14:54:13    |
      From: arne@vajhoej.dk              On 11/11/2025 10:56 PM, Lawrence D’Oliveiro wrote:       > On Tue, 11 Nov 2025 18:56:53 -0500, Arne Vajhøj wrote:       >> HA is about whether the system can continue to serve users in case part       >> of a box or an entire box fail - 24x7 vs 16x5 is about architecture.       >       > High availability is measured in “nines” -- e.g. five nines, six nines       ...       > even seven nines.       >       > How do big enterprises (like Google) achieve that? By not using       > mainframes. They set up data centres full of off-the-shelf PC hardware --       > one article I remember from over a decade ago said that Google, at that       > time, had 460,000 servers.       >       > All the hardware is obtained as cheaply as possible, except one component:       > the power supply. They buy quality for that, for power-efficiency reasons.       > As for the rest, it doesn’t matter if a box falls over every minute, or a       > hard drive crashes every few minutes; they have higher-level redundancy       > and recovery procedures that can routinely recover from all those       > failures, without the users ever noticing.       >       > No mainframe can match that.              Of course mainframes can match that.              The fundamental mechanism is the same for mainframes and       let us call it modern distributed environments.              You need N systems running to handle load. There is       a probability Pd of one system becoming unavailable.       You want Pr probability of handling the load.              You can calculate how many systems M you need to       achieve that.              N is smaller, Pd is smaller and the cost of a       system is much bigger for mainframes than for       x86-64 servers.              But the formula is the same. You can do the math.              IBM mainframes use OS clustering (like VMS) called       SysPlex. The modern distributed environments use       pure application level clustering. But that is       the "how" not the "what".              Arne              --- SoupGate-Win32 v1.05        * Origin: you cannot sedate... all the things you hate (1:229/2)    |
[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]
(c) 1994, bbs@darkrealms.ca