Re: Sorts (revised)

From: PAUL RAULERSON (pkraulerson_at_verizon.net)
Date: 06/27/04

  • Next message: PAUL RAULERSON: "Re: Is it possible to use the value of the PROGRAM ID within the source code?" 
    Date: Sun, 27 Jun 2004 16:51:58 GMT

    "Robert Wagner" <robert.deletethis@wagner.net> wrote in message
    news:40de91a6.155419779@news.optonline.net...
    > "PAUL RAULERSON" <pkraulerson@verizon.net> wrote:
    >
    > >Which specific supermarket chain might you be talking about?
    >
    > The company was Furr's, based in Lubbock. It no longer exists.
    >

    I thought Furrs was defunct cafeteria style fast food joint. :)

    > > The guy I
    > >currently work for did some
    > >*amazing* things at a Texas supermarket chain about the size you are
    talking
    > >about, and what's
    > >more, when he left, the company did about the same thing you describe.
    >
    > Anybody I might know?
    >

    Perhaps, his name is Bob Rainey but he was based more of Tyler. Guy really
    has a lot of talent when it comes to
    wringing performance out of systems. He also has the perhaps arguable credit
    for being the first person to
    deploy a major system on S/36 machines in RPG *without* using the RPG cycle.
    He now has a bit of a distaste
    for anything S/36'ish. <grin>

    > >By the way, I am familiar with HEB, and their IT budget includes a lot of
    > >things you didn't mention,
    > >like distributed systems in the store, warehousing, least cost routing
    for
    > >the trucks, and RFID. They
    > >do all that with a mainframe and keep the costs really low.
    >
    > We had all that too. I forgot to mention them. One of my favorites was a
    > speech-oriented Order Entry system that ran on PCs and received orders
    from
    > Telxon handhelds. Nowadays you can buy the technology from Dialogic et al.
    Back
    > then all we had was A-to-D converters. I wrote an audio editor in compiled
    Basic
    > using Bascom. Each MSDOS PC could process three parallel threads via
    > 'multi-tasking within the application'. In production, we ran 6-8 PCs.
    >

    I'm surprised - I worked for McLane corp and they are on the bleeding edge
    in some
    warehousing technologies, and RFID is rather new to them, as in the past 4
    years or so.
    Typical sales in the $16 billion dollar range, and about 20 warehouse
    locations all
    servered from corporate headquarters here in Texas, in real time. Been
    trying to do away
    with the mainframe, unsucessfully I might add, for about a decade now.

    > The most logic-intensive application turned store orders into warehouse
    picking
    > documents. It HAD to run in a narrow time window between midnight and 4am.
    It
    > did rationing for items with insufficient stock. It calculated Picking
    Standards
    > based on physics of tugger movement and warehouse geography. There were
    tons of
    > logic in the process. When it cratered, seldom, I personally did
    production
    > support. It was mission-critical.
    >
    > Earlier I worked at Kimbell in Fort Worth, which operated a showplace
    automated
    > warehouse. My system picked some items, especially produce, directly off a
    > receiving dock. It told suppliers on the purchase order "Ship to arrive at
    > receiving dock x at 3:00 pm." The produce warehouse had less than a one
    day
    > supply (vs. Furr's nine day supply). Kimbell was acquired by Winn-Dixie,
    in
    > whose hands it died a slow death. Last I heard, they still own it. Store
    count
    > is down from 800 to less than 50.
    >
    > >Also, I have yet to see a UNIX or Windows solution anywhere *near* as
    > >efficient of machine and
    > >programmer resources as a well designed mainframe application - which may
    > >consist of hundreds
    > >or even thousands, of programs. And that from a person who loves UNIX and
    > >loaded his Vax11/730
    > >with BSD as soon as he unpacked his personal tapes.
    >
    > It was basically a VSE shop augmented with PC applications. We weren't
    much into
    > Unix because I found it 'cryptic'. Nowadays, Unix is what I do. I've
    learned to
    > love it.
    >
    > Back in the day (early '80s) I ran drag races between the VSE mainframe
    running
    > VSAM (on FBA) vs. a 4.77 MHz PC running Realia's file system. The PC won
    .. to
    > my amazement.
    >

    Well, some of my buddies are suspicious of VSAM to this day - because they
    found it slower than
    ISAM/BSAM under VSE 20 to 25 years ago. But those same people have no
    experience at all with
    OS/390 or z/OS, and no feel for the scalability or performance of todays
    machines. Just take a look
    at the difference between say, RAMAC and an ESS-800. The ESS-800 typically
    has 4 four-processor
    gigahertz RS/6000 machines in it to try and keep up with the I/O load of a
    typical mainframe. Of course,
    part of that is redundancy and load sharing, but the statement is still
    basically true.

    The same machine can handle hundreds, perhaps even a thousand or so, PC's
    connected via Fibre
    and not even really break a sweat. That is a good indicator of the really
    VAST differences between
    a PC and a mainframe. A pSeries box almost always 'beats' an equivalent PC
    in any kind of a 'drag'
    race by the way, and even those machines are easily ovewhelmed by the sheer
    I/O capacity of
    a mainframe.

    > Nowadays I work with databases containing millions of rows. Mainframe and
    Unix
    > servers have similar CPU and disk speeds. The only difference is cost ..
    > mainframe cost is out the gazoo. Unix servers are just as fast for a
    fraction of
    > the cost, especially if they're running Informix rather than the 'safe
    choice'
    > Oracle, whose optimizer doesn't work right.
    >
    > Hand-wavers like to say X, the object of their derision, "isn't scalable".
    > Everyone nods in agreement, as though "scalable" were universally
    understood and
    > accepted. That's BS. I saw it at Sears, where the payroll system had
    performance
    > problems with Informix handling 400K employees. Then I went to US Dept of
    > Education Direct Loan Program, where Informix routinely handled tables
    with 30M
    > rows with no problem.
    >
    > Later, I worked on Major Big Time databases containing billions of rows.
    > Performance was not a problem under Oracle, Sybase OR Informix. It's all
    in the
    > way databases are indexed and tuned. Today, I saw runtime reduced from 1
    hour to
    > 1 second by applying indexes to an Oracle database.
    >
    >
    > >"Robert Wagner" <robert.deletethis@wagner.net> wrote in message
    > >news:40ddf3d4.115012207@news.optonline.net...
    > >> riplin@Azonic.co.nz (Richard) wrote:
    > >>
    > >> >robert.deletethis@wagner.net (Robert Wagner) wrote
    > >> >
    > >> >> >>In the six shops I managed for 20 years, all programmers were in
    the
    > >upper
    > >> >> >> 10% skill-wise.
    > >> >
    > >> >> That speaks to quantity. Quality is subjective. In my Humble
    Opinion,
    > >their
    > >> >> code was beautifully crafted.
    > >> >
    > >> >> The following passage was deleted. Using 'The Way Things Are' as a
    > >working
    > >> >> interpretation, you argue for the status quo, which is piss poor
    Cobol.
    > >> >
    > >> >I have noticed a common theme. It seems that everything that you have
    > >> >been involved with is the 'best', is beautiful, is skilled, while
    > >> >everything that you have never even seen is 'piss poor', mediocre and
    > >> >'a travesty'.
    > >>
    > >> I encourage programmers to strive for beauty.
    > >>
    > >> >> If you want more than one-line responses, post something more
    > >thoughtful than
    > >> >> attacks on my personality.
    > >> >
    > >> >He is not 'attacking your personality', but is commenting on your
    > >> >observable _behaviour_. You _did_ call what mainframers did as 'a
    > >> >travesty' and did admit to calling them mediocre and to baiting them,
    > >> >even when you have never seen any of their actual code.
    > >>
    > >> I've seen millions of lines of their actual code. I've rewritten
    thousands
    > >of
    > >> their programs, in some cases every program in the shop.
    > >>
    > >> The first thing I do is to modularize IO, one program per file. You
    talked
    > >about
    > >> doing the same. This allows redesign of file structure, or moving it to
    a
    > >> database, without recompiling or testing application code. When doing
    > >that, I
    > >> usually regression test one or two systems.
    > >>
    > >> The result was modular code that was much easier to maintain and, more
    > >> importantly, make changes with confidence they'd work right the first
    > >time.
    > >>
    > >> For instance, I worked at a medium-sized supermarket company in Texas
    that
    > >had
    > >> 200 stores and 2 warehouses. Supermarket industry norms said we should
    be
    > >> spending .5% of sales on IT. The company's sales varied between $1B and
    > >$2B (the
    > >> higher number after a major acquisition), so we should have been
    spending
    > >> between $5M and $10M. That translates to about 50 programmers and
    > >analysts.
    > >> Several competitors I knew details about, such as HEB in San Antonio,
    did
    > >indeed
    > >> spend that much and have that many programmers. We spent $1M per year
    on
    > >total
    > >> IT (hardware, software, paper and programmers) and did it with 6
    > >> programmer/analysts. The difference was $4-9M hard cash to the bottom
    > >line.
    > >> After I left, the IT budget went from $1M, where it had been for 7
    years,
    > >to
    > >> $7M, within the industry norm. Performance actually decreased and
    > >programmers
    > >> said it was no longer a fun place to work.
    > >>
    > >> Anticipating rebuttal, you'll say my methodology was tied to a 'cult of
    > >> personality' rather than a universal solution. Not so.. After I left,
    they
    > >> discarded my systems and replaced them with a canned ERP solution.
    > >>
    > >> >This _may_ be a result of your personality, or may be because of a
    > >> >lack of it, I could not judge that, I can only judge what I see in the
    > >> >words.
    > >>
    > >> I've never had trouble getting along with a good programmer, except
    here.
    > >My
    > >> enemies, except here, have been mediocre and bad programmers, who fealt
    > >> threatened.
    > >>
    > >> When I was manager, the problem solved itself because the bad ones quit
    > >before I
    > >> fired them. In two cases, they tried to make me look bad by hacking
    into
    > >my
    > >> programs from outside and sabotaging them. In both cases I was able to
    > >capture
    > >> an image of their 'tool' and disassemble it. When confronted with
    tangible
    > >and
    > >> irrefutable evidence, they quit.
    > >>
    > >> Why is my reception in CLC discrepant with it in the real world? I
    didn't
    > >live
    > >> in a niche or sheltered environment. I worked with people from divers
    > >> backgrounds ranging from CS academics to self-taught geeks with low
    social
    > >> skills. The answer must be in the dynamics of electronic communication
    vs.
    > >> face-to-face.
    > >>
    > >> Robert
    > >>
    > >
    > >
    >

  • Next message: PAUL RAULERSON: "Re: Is it possible to use the value of the PROGRAM ID within the source code?"