From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=unavailable autolearn_force=no version=3.4.4 X-Received: by 2002:a6b:815f:: with SMTP id c92-v6mr8017600iod.22.1524348695182; Sat, 21 Apr 2018 15:11:35 -0700 (PDT) X-Received: by 2002:a9d:70d7:: with SMTP id w23-v6mr334703otj.3.1524348694912; Sat, 21 Apr 2018 15:11:34 -0700 (PDT) Path: eternal-september.org!reader01.eternal-september.org!reader02.eternal-september.org!feeder.eternal-september.org!news.uzoreto.com!weretis.net!feeder6.news.weretis.net!feeder.usenetexpress.com!feeder-in1.iad1.usenetexpress.com!peer02.iad!feed-me.highwinds-media.com!news.highwinds-media.com!f63-v6no2028519itc.0!news-out.google.com!15-v6ni2615itg.0!nntp.google.com!f63-v6no2028516itc.0!postnews.google.com!glegroupsg2000goo.googlegroups.com!not-for-mail Newsgroups: comp.lang.ada Date: Sat, 21 Apr 2018 15:11:34 -0700 (PDT) In-Reply-To: <877ep02qru.fsf@nightsong.com> Complaints-To: groups-abuse@google.com Injection-Info: glegroupsg2000goo.googlegroups.com; posting-host=47.185.233.194; posting-account=zwxLlwoAAAChLBU7oraRzNDnqQYkYbpo NNTP-Posting-Host: 47.185.233.194 References: <1c73f159-eae4-4ae7-a348-03964b007197@googlegroups.com> <878t9nemrl.fsf@nightsong.com> <87h8o7lowg.fsf@nightsong.com> <8736zqkwat.fsf@nightsong.com> <6839088c-f221-4650-a6ea-1841ae539486@googlegroups.com> <87zi1xu0xg.fsf@nightsong.com> <877ep02qru.fsf@nightsong.com> User-Agent: G2/1.0 MIME-Version: 1.0 Message-ID: <256adb42-6bdf-4bee-a137-a7d69ca2daa6@googlegroups.com> Subject: =?UTF-8?B?UmU6IEhvdyB0byBnZXQgQWRhIHRvIOKAnGNyb3NzIHRoZSBjaGFzbeKAnT8=?= From: "Dan'l Miller" Injection-Date: Sat, 21 Apr 2018 22:11:35 +0000 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Received-Bytes: 6592 X-Received-Body-CRC: 3667333160 Xref: reader02.eternal-september.org comp.lang.ada:51664 Date: 2018-04-21T15:11:34-07:00 List-Id: On Saturday, April 21, 2018 at 4:28:23 PM UTC-5, Paul Rubin wrote: > Simon Wright writes: > > What about the bank keeping track of your money? >=20 > Not realtime, or at least not hard realtime. Nobody cares if checking > your bank balance online takes a millisecond longer than it was supposed > to. They might care about the statistical distribution of the latency > but they don't care at all about a single query. (Hard) realtime means > that disaster ensures if a deadline is missed even once. https://en.wikipedia.org/wiki/Erlang_(unit) Hard realtime has fixed-duration time deadlines that if missed make the lat= e numerically-correct answer or late logically-correct event just as wrong = as if the answer was numerically incorrect or illogical. Elastic realtime = often can be nearly as strict: if the rate of successfully-computed retire= ment* of transactions cannot keep up with the erlangs of arrival rate of co= mmencement of non-elidible transactions during extant-expected-peak (not th= eoretical peak) busy periods whose duration is longer than the elasticity t= ime-duration, then some of those incoming transactions are going =E2=80=9Cf= all on the cutting room floor=E2=80=9D so to speak=E2=80=94the proverbial b= it bucket. If those transactions are Visa or MasterCard (or even Bitcoin n= owadays) financial transactions, societal pandemonium will occur far more s= everely than one 90-year old grandma dying because her pacemaker skipped on= e beat. If Visa and MasterCard cannot successfully account for a slew of t= ransactions during extant-peak-busy times, suddenly people stop trusting el= ectronic forms of money as a stand-in for paper & metal-coin money. It wou= ldn't take too many steps to suddenly quickly arrive in points of freeze-up= of the financial system to the point of, for example, General Motors Corpo= ration going out of business because they cannot make payroll because they = cannot take out their rolling 30-60-90-day rolling line-of-credit loans. (= General Motors Corporation was replaced by General Motors, Inc. for substan= tially this reason, eliding details to make it fit succinctly herein.) * retirement here would (must) be one of the following: 1) sufficient funds, payment made; 2) insufficient funds, payment denied; 3) DoS/DDoS prevention, transaction balked overtly. Successful retirement prohibits silent failure where retried transactions (= e.g., due to timeout of the momentarily prior same-terminal/same-card trans= actions), especially where the tries and their retries (e.g., due to custom= er/merchant frustration) get confused or are non-idempotent.=20 So maldesigned systems where erlangs of transaction-commencement exceed erl= angs of successful retirement are another form of (slightly elastic**) real= time failure due to results arriving late to the point that they are comple= tely useless/wrong. ** The elasticity time duration is the time-duration of a timeout of a cred= it-card customer an a terminal, because retries out of frustration add furt= her traffic during the would-have-been-peak time period, raising the peak t= o disastrously higher and higher levels, assuring further deadline misses a= s retirement rate cannot keep up with transaction commencement rate. This is how network processors define realtime regarding arrival rate of, s= ay, Ethernet frames at, say, 10Gb/s. The elasticity time duration is so sh= ort that elastic realtime very closely approximates hard realtime the way t= hat engineering zero due to tolerance approximates true zero that the diffe= rence between definitions is purely academic and effectively moot. > > What about those crazy high-frequency traders? >=20 > Their machines are not in data centers. The heck they are not! Arbitrage traders place their servers as close to t= elecom data centers (called central offices) as they can get! At any price= ! If they cannot get into the same telecom central office as the target fi= nancial market, they might instead even be in an outdoor cabinet along a ra= ilroad easement so that they can be at a repeater station along a main-trun= k fiber-optic cable transmission line just outside their target financial e= xchange to shave off a few milliseconds of router switching time here and t= here. Indeed, they try to eliminate layer-3 Internet routing entirely as m= uch as possible, leasing expensive layer-0 (e.g., DWDM wavelength) or layer= -1 (e.g., SONET-SDH) or layer-2 (PBB-TE Ethernet) transmission channel to h= ave as direct least-routed least-touched-by-hardware-dataplane switch-fabri= cs as possible! Arbitrage trading is a slightly different definition of =E2=80=9Clate=E2=80= =9D. Instead of fixed length of time being the deadline, the deadline is b= eing the fastest. Second in the race misses the deadline, and is considere= d total failure, because the arbitrage opportunity was lost (along with hun= dreds of thousands or millions of dollars).