History, reasoning, and anything else notwithstanding, charging for something that all other hardware does for FREE is hard to justify to a bean counter (or for that matter, most reasonable people) Ed used the correct word. Embarrassing. > -----Original Message----- > From: IBM Mainframe Discussion List <[email protected]> On > Behalf Of Alan Altmark > Sent: Friday, May 14, 2021 9:53 PM > To: [email protected] > Subject: Re: STIMERM LT value > > On Fri, 14 May 2021 05:28:52 -0700, Ed Jaffe > <[email protected]> wrote: > >Reasonable people can disagree about what the word "pricey" means, but > >no one can defend the indefensible. > > > >Every Intel-based Linux, Windows or Mac, at every hardware price-point, > >can synchronize its clock with an external NTP sever in real-time > >without a reboot. > > The lineage and history of the clocks and time on microprocessors is very > different than that of the mainframe and leads inevitably to the OS needing > to use NTP and have the OS be the Arbiter of Time. (Do apps call > gettimeofday() or do they read the timestamp register? They call > gettimeofday().) > > On the mainframe, we allowed (encouraged?) applications (subsystems, > user apps, middleware) to use the TOD clock. STORE CLOCK is an > unprivileged instruction. And the tradition predates the introduction of leap > seconds, so everyone calculated the same time using the same algorithm > over and over and over and over again. If you have someone doing STCK, > but not using CVTLSO, they're likely to calculate the wrong time. At the end > of the day (no pun intended), the TOD clock needs to be in sync with UTC. > And once that decision is made, it becomes an obvious waste of resources to > implement an ntp client in the OS. > > With ntp, if multiple servers are all synced to the same time source then they > are, by acclamation, synced with each other. But that's not good enough > since there are lots of places with TOD clock or TOD clock-based values. > They need to be on the same time line. > > >For the high six-figure price we paid for our z15, it should have come > >with similar functionality built-in, but didn't. > > > >IBM's current positioning in this regard, as Dave Gibney and others > >discovered, is a source of embarrassment for the platform. > > Help me with this, Ed. Did you receive a proposal that didn't include STP? > Or > at least offer it as an option? If it wasn't offered, that bothers me. IMO, > all > proposals should include cryptos (where not prohibited by law), a pair of > copper OSAs, six or more fiber OSAs (speed and type customer choice), a > pair of RoCE cards, STP, STP links, and a pair of HMCs (unless you already > have them). If it's an IFL box, then add some FCP adapters. Getting this > stuff in the initial sale is going to be cheaper than adding it later. > > Then if the client wants to reduce the price by redlining items, that's on > them, but I think IBM or the BP should be bringing a full function proposal. > > Alan Altmark > IBM > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN
---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
