Path: csiph.com!eternal-september.org!feeder.eternal-september.org!nntp.eternal-september.org!.POSTED!not-for-mail From: Peter Flass Newsgroups: comp.os.linux.misc,comp.unix.shell,alt.folklore.computers Subject: Re: NYT and Stratus (Re: [OT] Dead DEC Punch card times) Date: Sun, 8 Nov 2026 20:33:45 -0700 Organization: A noiseless patient Spider Lines: 26 Message-ID: <11creup$trgi$1@dont-email.me> References: <18dee4b6cab427fa$865$2574533$802601b3@news.usenetexpress.com> <11cdg33$47sq$1@dont-email.me> <11cdp6v$1qa8$3@gal.iecc.com> <875wy85ba3.fsf@localhost> <11cpv8t$c3gs$1@dont-email.me> <11cqpn5$nfgo$2@dont-email.me> <11cqqh2$ntcu$1@dont-email.me> <11cr1ub$q796$1@dont-email.me> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit Injection-Date: Mon, 09 Nov 2026 03:33:46 +0000 (UTC) Injection-Info: dont-email.me; logging-data="978450"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX193G+5Ddm+72dOasYyWdQmU"; posting-host="0d82f543747792d1f39ffbaec9941121" User-Agent: Mozilla Thunderbird Cancel-Lock: sha1:zWWUs1Qn2Fqbc2+QU3NpWCtnl9k= sha256:wYwUJe1TvdYGaHU22Kao2TMaI8tZcilwi3elTvT88cI= sha1:j3CfnNQnl2IpDPl7BLqKWiRuT5E= sha256:9V8nxIXByb/VZt6EYQynscb9oJOFcbB3t77jgpkZ/r0= Content-Language: en-US In-Reply-To: <11cr1ub$q796$1@dont-email.me> Xref: csiph.com comp.os.linux.misc:95374 comp.unix.shell:27351 alt.folklore.computers:236682 On 11/8/26 16:51, Lawrence D’Oliveiro wrote: > On Sun, 8 Nov 2026 13:45:03 -0800, Lars Poulsen wrote: > >> On 11/8/2026 13:31, Lawrence D’Oliveiro wrote: >>> >>> There was a paper on how Google did it, published 10 or 20 years >>> ago. At that point, they had a total complement of about 460,000 >>> x86 servers. They would buy the cheapest component they could for >>> every component but one: the power supply. >>> >>> With that kind of scale, who cares if 100 or 1000 machines are >>> broken at any moment? The users certainly won’t notice. >> >> The parallelism of a Google is a very different beast from the >> robust parallelism of a banking mainframe. > > Mainframes have never been that “robust”. You’ve mentioned elsewhere > learning about Stratus and Tandem, who took robustness to a new level > back in the 1980s, beyond anything the mainframe world could match. > Think of that kind of redundancy multiplied tenfold, a hundredfold, a > thousandfold, all built on open-source technologies: that’s the kind > of world we live in today. > Mainframes typically achieve an uptime percentage between 99.999% (five nines) and 99.9999999% (nine nines)