A (mostly) daily journal of the
trials, tribulations, and random observations of Robert Bruce Thompson, a
writer of computer books.
Monday,
November 16, 1998
Well, I ended up spending most of yesterday working on the
web site. Originally, I'd just intended to reformat existing pages using
the new format. As I worked on them, however, I ended up revising a lot of
existing text and adding a lot of new text. I didn't keep track of what I
worked on, so if you want to read the new and revised stuff, you'll just
have to find it. At least each such page is identifiable by being in the
new format and having yesterday's date on it. Sorry.
* * * * *
On Saturday, I'd commented that the BackupExec verify pass
couldn't be doing an actual compare against the contents of the backup
tape it had just created, and speculated that it was calculating a CRC
value for each file as it backs it up, storing that CRC with the file,
recalculating the CRC from the tape data during the verify pass, and then
comparing the two CRCs. I got the following mail last night from Gary M.
Berg:
I have had the same experience as you with
Backup Exec; it reads the tape during the verify pass and doesn't
compare it to the source. I believe it does actually do CRC/Checksums or
something on this pass, and not just read every byte. I suspect there is
enough documentation on their web site to confirm this, and there may be
enough in the manual. In fact, here's the info from their site:
http://www.seagatesoftware.com/bewinnt/
Thanks, Gary. I figured it was something like that, or at
least I was hoping it was. Doing it that way makes sense when you think
about it. I appreciate you taking the time to find out for sure and let me
know.
* * * * *
I also got another email yesterday that deserves comment.
I won't reproduce that mail here. It was the usual semi-literate
nastygram, this time accusing me of being racist (!) on the basis of what
I'd written in The U. S. Open and Big
Servers. Huh? I made three basic points in that article: (1)
professional tennis players are missing an opportunity by hitting a
"second serve" instead of a two "first serves." (2)
the Cyclops device that they use to call lines cheats, and (3) the speed
gun that they use to measure the speed of serves overstates the speed of
women's serves relative to that of men's. I'm not sure how anyone could
get "racist" out of that article. Perhaps I missed something.
* * * * *
And this mail, from Tim Werth:
After reading your question of NT throughput
over a network this morning I noticed in John C. Dvorak's column
for PC Mag that he has heard about throughput problems w/NT. Thought
that was kind of an interesting coincidence. L8r
I've been aware for quite a while that NT sometimes has
problems saturating a 100BaseT network, as opposed to NetWare, which
doesn't. In fact, Microsoft always does comparative testing in a switched
environment to minimize its throughput differential relative to NetWare.
I'd never even suspected, however, that it might not be able to saturate a
10BaseT network, and this using name-brand (3Com) NICs and a 3Com hub that
are connected with Cat 5 cable. If that turns out to be the case, NT is
pathetic. Thanks for your note.
And this mail, from Christl Buskohl:
After reading your page
on the Pentax Honeywell, I had to ask. I have one, and need to replace
the battery (the battery only runs the light meter which could be why
yours isn't working) and can't find anyone who has a clue as to what the
battery would be. Even showing it to some people, they don't have a
clue. Any idea on what kind or voltage of battery these cameras need for
the light meter? Any help will be appreciated.
Although I lost track of the manual for this camera a
decade or more ago, I'm almost certain that it uses the standard Energizer
EPX76 or equivalent silver oxide hearing aid battery that you can buy in
any drugstore. At least that's what I've been using. I wish the problem
was the battery, but the meter is dead even with a fresh battery. I have a
copy of Herbert Kepler's book The Honeywell Pentax Way around here
somewhere. I'll see if I can locate it and find out if it mentions the
battery type. I'll also post your message and my reply in case anyone else
knows.
Ordinarily, I don't publish email addresses because a lot of people,
fearing spam, have asked me not to. However, Cristl specifically requested
that anyone who knows for sure what battery type the Honeywell Pentax
Spotmatic requires reply directly to cbuskohl@wwcc.cc.wy.us.
Tuesday,
November 17, 1998
If it's not one thing, it's another. This morning, about
12:30, I had just put my book aside and turned off the bedside light when
the UPS in Barbara's office let out a single loud beep. Okay, I figured,
we've been having numerous momentary power failures lately, so we must
have just had another one. A half hour or so later, it let out another
beep, and a couple minutes later, still another. I was still awake when
these happened, and didn't notice any flickers in the night light across
the hall. At that point, I started to wonder if the problem was with the
UPS itself.
This is the same APC BackUPS-600 that I moved to Barbara's
office a few months ago. It needs a replacement battery (it only provides
about a minute of run time at minimal load), so I thought perhaps what I
was hearing was a dead battery warning. I went in and shut down thoth
and turned off the UPS so we could get some sleep. When I got up this
morning, I turned the UPS and thoth back on, and there haven't
been any chirps so far, so perhaps we really were having momentary power
outages.
I guess what I should do is just order a replacement
battery, although what I'd like to do is go down to Sears and get a
deep-cycle marine battery to use instead. There wouldn't be any convenient
way to connect it, though. At 600 VA, that thing needs to pull 50 amps net
out of a 12 V battery. That'd mean running what amounts to a set of heavy
jumper cables between the battery and the UPS. I guess I'll stay simple
and just buy a replacement battery. Also, I seem to remember that there
are some DIP switches that allow you to set options for the alarm. I'll
have to dig out the manual and check those.
I spent some time last evening revising still more pages
and adding/editing text. At this rate, I'll have a whole new site in the
next month or so.
* * * * *
And this mail, from Maurice McDonell:
Found you via Pournelle's site, I was just
about to order an Olympus DL 320L for a Christmas gift (spouse)when I
chanced on your article on that
Sony Mavica. It happens that she has actually handled one of those
things; last year at Fry's in Sacramento. It was the only one on display
(of 20 or so) that worked and some factory folks were there to charm her
- she loved it. I thought that the floppy output was a great idea. She
worried about the 640 x 480 resolution.
Called around, including New York camera
shops - they hated it - unreliable, non-photograph photography or some
such thing, very short warranty, etc. We waited. Last week, a B&H
camera catalog arrived by USPS. It has been reorganized by category,
first chapter is digital photography. I was beginning to think that an
intersection of the same product in a photo mag and in PC Usermag might
be enough to stimulate a purchase. Thanks to your commentary and
samples; I could evaluate the images (nice Dog). On the screen (17"
KDS) they looked great; printing is lousy (HP 670@ work, 820 @ home).
Spouse got a "new" camera last
Winter - Pentax K-1000 - at last she can take pix without fighting the
settings. Good photos, she takes, now. I scan (HP Photosmart) them for
future use somewhere. They make great viewing on the screen as .BMP
files. I do not know what to do about hard copy output. Maybe send the
best ones to a service bureau ????
Your review was crisp and easy to
understand. I have copied it for spousal information.
I have not seen the term "30" in
use since UCLA won all their basketball games in one season - I believe
Jim Murray of the L.A. Times used that term to describe the season.
Thanks for your help.
I'm glad you enjoyed the article, and thanks for the kind
words. My one gripe about the Sony was that its maximum resolution was
only 640X480. I noticed an ad in PC Magazine the other day that announced
that Sony now has a 1,024X768 version of that camera out. As far as I'm
concerned (assuming the price isn't too outrageous), that tips the scales
in favor of the Sony versus any of its current competition. Having that
floppy disk drive really makes life a lot easier.
I should probably update that Mavica article. I'm kind of hoping I can
sweet-talk Sony out of an eval unit of the new 1024X768 version of the
Mavica. If not, I may have to buy one. I went to the B & H Photo site
to check the price on the new Sony camera. I found two 1024X768 versions.
The FD-81 costs $720 street, and has a 3:1 optical zoom. The FD-91 costs
$1,000, and includes an upgraded 14:1 optical zoom lens (the equivalent of
a 37mm to 500+mm zoom on a 35mm camera), a better viewfinder, and a couple
of other minor improvements.
Wednesday,
November 18, 1998
Spent most of yesterday cranking away on the new O'Reilly
book and trying to sweet-talk vendors out of eval units, with less than
the usual success. David Rogelberg, my agent, called late yesterday
afternoon to say that the folks for whom we had been doing the white paper
wanted to do the right thing by paying us for our efforts. They asked him
what he thought was fair, and he told them that their initial $20,000
payment was about right. They thought that was way too much, but agreed to
pay $5,000 as a nominal acknowledgement of our work. Divided two ways and
after deducting David's share, that's $2,125 each. That for close to 200
hours of work over 14 straight days. That comes out to about $11/hour
calculating at straight time, or about $8.50/hour figuring time-and-a-half
for overtime. Not slave wages, but certainly nowhere near normal tech
writing rates. Still, two grand is a whole lot better than the proverbial
sharp stick in the eye.
* * * * *
And I got this mail last night from Roger G. Smith regarding my idea to
replace my APC UPS battery with a deep-cycle marine battery. My comments
are embedded:
Bob, if you figure out how to do this, let
us know. I've thought about it, but you have to make some provision for
venting or neutralizing the battery acid fumes. You may have had
experience with the corrosive effects of those fumes with a battery just
sitting idle. A big lead-acid marine battery on a trickle charger could
produce real problems in an office.
Well, acid fumes wouldn't worry me much. But charging a
lead-acid storage battery generates hydrogen, and I smoke. Nothing like a
hydrogen explosion to disrupt one's peaceful enjoyment of a pipe of fine
tobacco. Actually, that probably wouldn't be a problem. I don't know what
the explosive concentration of hydrogen is, but I'd guess it'd have to be
several percent at least. There's no way a small battery is going to
generate that in a 200 square foot room, unless the door and windows were
sealed through several full charge/discharge cycles. I've got an original
copy of Vinal's book on storage batteries around here somewhere. It was
written in about 1910, has been revised several times since then, and is
still a definitive guide to lead-acid batteries as far as I know.
The battery could be remotely located, but
running 30 to 60 feet of _large_ cable does not appeal. Remote location
of the UPS has its own set of problems, though I suppose you could take
out a small home improvement loan and have one centrally located Clary
with 2 kW generator backup. Then run new AC wire for the protected
circuits.
Actually, what I was thinking of doing was locating the
whole UPS remotely and running 10 or 12 gauge Romex to it. I don't have a
table handy, but I'd guess that running 60+ amps over 60 feet would
require something like 4 or 6 gauge cable to get anything like a
reasonable voltage drop. I figured I'd install a run of Romex with a
standard receptacle at each end and make a male-male backfeed wire to join
the remote UPS to the receptacle on that end. But the truth is that I
don't have time to mess with this. It's a lot easier just to buy a $50
replacement battery and have done with it.
It makes a certain amount of sense to be a
warm, well fed writer with electricity during, say, a 10 day ice storm,
but, yee gads! With LAN monitoring, a remotely located UPS is more
practical. If you get any useful ideas, be sure to write about it.
Yes. I have a 6 kV Generac sitting in the basement right
now, along with two 6-gallon fuel cans and a bottle of gasoline
preservative that Barbara picked up at Wal-Mart. Now all I need is fuel,
oil, and some means to connect it to the stuff I want to run. I'll
probably write about all this stuff at some point, but for now, Good
Enough is Good Enough. My wife said I was nuts to do anything except order
a replacement APC battery. "Pay the man the two dollars" is
usually good advice.
* * * * *
And now, a question. Am I cheating by publishing so much
mail from readers, or do you like seeing it? I get a fair amount of mail
each day, and I'm starting to publish more of them. Many aren't
appropriate for publication. Not that there's anything wrong with them.
They're just not of general interest. But I think publishing some of the
interesting ones helps fill in the blanks. Right now, I'm cranking on my
O'Reilly book, and it gets pretty boring for me, and presumably for you,
if I just report that I wrote X number of words yesterday.
I can't even be specific about the title or topic of the
book. All publishers, not just O'Reilly, want to keep forthcoming books
secret. Pournelle let the cat out of the bag on his web site by reporting
briefly that the two of us were going to do a big hardware book together.
It wasn't Pournelle's fault--no one asked him not to mention it. From
comments I've made and essays I've written, it's probably not much of a
stretch to guess the topic of the book I'm working on alone right now,
which kind of dovetails with the joint book Pournelle and I want to do.
Still, O'Reilly would probably not be happy to have us discussing it in
detail on a web site. If I happened to mention that I'd been writing today
about, say, ATA/IDE and PIO/DMA modes, they'd probably think I was giving
too much away. They like to surprise the competition, and I kind of see
their point.
I, on the other hand, figure the more people that know,
the better. It's not like the competition can do anything to stop us. I
like the approach the Green Bay Packers used back in their 60's glory
days. You knew exactly what they were going to do. They knew you knew
exactly what they were going to do. They did it anyway--ran the ball right
down your throat, and there wasn't a damn thing you could do about it.
I think that being up-front about what we're doing, at
least with the readers of our web sites, has some very real advantages.
One suggestion I made was that the readers of Pournelle's web site would
make ideal technical reviewers for our book. Jerry was very enthusiastic
about the idea, and O'Reilly is cautiously optimistic Obviously, they
don't want the text of our book making the rounds on the Internet--neither
do we--but that's a far cry from circulating draft chapters among a small
group of honorable people.
Getting a good tech review is tough, because it pays
little or nothing. Usually what ends up happening is that O'Reilly authors
take turns tech reviewing each other's books gratis or for a small
honorarium. For example, AEleen Frisch, author of Essential
Windows NT System Administration and other O'Reilly books, must
have spent literally a week doing her tech review of Windows
NT TCP/IP Network Administration for Craig Hunt and me. I don't
know what O'Reilly paid her to do that, but I'd guess it was $200, if
that. Not much for 40 hours or more of work. But in return, Craig and I
will joyfully do a thorough tech review on her next book, if she asks us
to.
At any rate, the real problem is that tech reviewing pays
a buck or two a page, if that, and there's no way anyone can do an
anywhere near complete job at that rate and still eat every day. A lot of
times, the tech reviewers just get one or a few free books of their choice
or something similar. My idea was that instead of using a few (poorly)
paid reviewers, we solicit volunteer tech reviewers from among Jerry's and
my readers. I think we'd get quite a few volunteers and a very thorough
review by some real experts on the topics. We might have half a dozen
reviewers for each chapter, but any given reviewer might review only one
or a few chapters. What do you think?
* * * * *
Talk about brain-dead software. That term can't be
Politically Correct, nowadays, can it? Cerebrally-challenged, perhaps? At
any rate, I just ran the FrontPage98 spell checker on this document. It
choked on "WalMart" and suggested I replace it with
"Wal-Mart". I did so, but when I ran the spell checker again, it
got to "Wal-" and choked again. That's pretty bad. I take its
suggestion, and it immediately turns around and tells me that the spelling
it suggested is wrong. Geez.
And speaking of Microsoft software, I just heard that the
judge announced they lost--some interim challenge, no doubt. I mean, Bill
Gates' net worth increased by something like $20 billion last year. We
have a president who literally sells nights in the White House for
$100,000. How long can it be before Gates buys Clinton? For that matter,
how long can it be before Gates hires Iraq or Libya to do something about
his problem?
Hearing that Microsoft lost brings mixed emotions for a
lot of people, kind of like seeing one's mother-in-law drive over a cliff
in one's new Ferrari. Oops. That can't be PC either. I don't like some of
the things that Microsoft does, but they've done nothing to warrant the
attacks the federal government has launched. Since when has it been the
government's right to interfere with free enterprise? No one has argued
convincingly that Microsoft has violated any law whatsoever, and yet the
full force of the soi disant Justice Department has been brought to bear
on them.
Bad as this is, it's not as bad as what's being done to
the tobacco companies. Again, they've violated no law, and trial after
trial has found them blameless. No one has established causality between
cigarette smoking and the bad things that it's accused of causing. Even if
causality had been established, cigarette smokers smoke of their own free
will. I know. I smoked a pack or two a day for more than a decade, from
the time I was about 20 until I was in my mid-thirties.
And the addiction argument is a red herring. No one,
despite many sustained efforts to do so, has ever established that
cigarette smoking is addictive in any medical sense. They weasel-word and
call it "psychologically habituating", but that's a crock. I
know. I stopped smoking by accident. About ten years ago, I came
down with the flu and didn't much feel like smoking. After a week or so
(with no desire to smoke in the interim), I'd recovered from the flu, had
no desire for a cigarette, and never started smoking cigarettes again.
I actually went several weeks without smoking anything at
all, but I happen to enjoy smoking. So, I decided to start smoking a pipe.
I've been doing that ever since, not from any compulsion to do so, but
simply because I enjoy it. I also smoke a cigarette occasionally, usually
when I can bum them from my friend John Mikol. The last few years, I've
probably averaged one or two cigarettes per month, usually clustered into
several in a day followed by months of smoking none of them. At no time
have I felt any urge to run out and buy a carton of cigarettes.
At any rate, here we have a legal product, the consumers
of which use it voluntarily, and the government decides to bury the
industry. Not by something as straight-forward as criminalizing
cigarettes, no. That wouldn't do. After all, the government rakes in
billions of dollars on tobacco taxes, and they don't want to give that up.
So, instead, they force the tobacco companies to collude with government
in their own demise. They'll have to pay out more than 200 billion dollars
in so-called reimbursement for health care costs to the states, and have
been forced to curtail advertising their own legal product.
And the states' arguments are specious. They argue that
smoking cigarettes causes illnesses, and that their subsidized health care
programs must absorb additional costs caused by smoking. Disregarding the
facts that causality between smoking and illness has never been
established (really) and that not every one agrees that the states should
be subsidizing health care in the first place, that leaves one crucial
problem with their arguments. They are either right or wrong when they
argue that cigarette smoking causes additional and more severe illnesses.
If they are wrong, there is clearly no reason why the tobacco companies
should be paying billions to them. If they are right, the tobacco
companies still should not be paying those subsidies. Here's why:
Consider those smokers who died young from lung cancer,
heart disease, or whatever. The states had to pay for their medical care,
and that is the basis of the states' claims. However, if in fact
cigarettes are killing people early, then the tobacco companies are doing
the states a favor economically from an actuarial point of view. Not to be
cold-blooded about it, but each smoker who dies young saves the states
money in the long run. The states incur an up-front cost during a smoker's
final illness, certainly. But that cost pales compared to the expected
costs that the state would incur for ordinary medical care, pensions, and
so forth had that smoker lived to, say, 75 years old instead of 55. In
purely economic terms--and that is the basis of these states' claims--the
states are actually better off if cigarettes do kill people off young.
Any reasonable person would recognize the validity of this
position. You might argue the case morally, but not economically, which is
what the states have chosen to do. On that basis, they are persecuting
rather than prosecuting the tobacco companies, and they should be punished
for doing so. Incidentally, although I have never heard this argument made
publicly, I have been told by a high-level executive at one of the tobacco
companies that they proposed this defense, but that the judge refused to
allow them to put it forward. Some judge.
* * * * *
And this mail I got a month or so ago from a reader who
asked that his name not be published. It seemed particularly appropriate
to print it now, after all that stuff I just got through writing:
I love your journal and read it
every day. I wish you wouldn't write so much though. It takes too much
time to read it all.
Hmmm. There's a point in there somewhere, but
I'm not entirely sure what it is. Perhaps "good things come in small
packages." Thanks for the kind words, though. If you think you spend
too much time reading it, imagine how I feel about how much time I spend
writing it. In my defense, I write for the same reasons that a lot of
people watch TV or go for walks. I find it relaxing. That probably sounds
stupid, given that I write all day for a living, but writing for money and
writing for enjoyment are very different things. From your point of view,
the good news is that I'm very busy now, and about to become more so, so
the length of these essays will undoubtedly shrink over the next few
months.
* * * * *
And last night I decided to do something about the power
blips we seem to be getting several times a day now. They're not major
outages--just failures that last a small fraction of a second. Not even
enough to kill the digital clocks or dim the lights noticeably most of the
time. The only way I knew about some of them was that Barbara's UPS kept
squealing. Others lasted long enough to dim the lights briefly. This is
now happening eight or ten times a day. It affects all circuits, so it has
to be something outside the house, presumably a problem with the drop
cable from the street.
Last night at 9:05 p.m. it happened again, and I finally
decided to call Duke Power. I explained what was going on--with another
failure and UPS squeal in the middle of the call to emphasize the
problem--and the woman said they'd send someone out today. I figure a
squirrel has been gnawing on our drop cable or something.
At any rate, that left last night to worry about. I didn't
want that UPS alarm going off in the middle of the night again. I
remembered that there was a way to delay the alarm, but I couldn't find
the BackUPS-600 manual, of course. I hit the APC web site, and it said to
move DIP switch 1 to the up position to delay the alarm. I did that, and
we had another short blip a few minutes later. No alarm. Good. We got a
good night's sleep with no interruptions from the UPS.
Then, in the can't-win-for-losing department, I was
sitting here this morning writing away on the O'Reilly book when my desk
lamp died. I immediately saved off, started a shutdown on my main
workstation, and went roaring back to Barbara's office to shut down thoth.
(Why I don't connect the interface cable to the UPS and install the
automatic shutdown software, I don't know.) As I came into Barbara's
office, my shadow crossed the light sensor on the nightlight in her
office, and it came on normally. Oh-oh. I turned on her desk lamp and it
came on normally, too. Crap. I headed back for my office. Sure enough, the
light bulb in my desk lamp had just burned out.
Once Duke Power figures out and fixes the problem, I can
go back to having the UPSs beep when the power fails. Better still, I'll
install PowerChute and let it shut things down automatically.
Thursday,
November 19, 1998
After yesterday, I'm written out for a while. Not only did
I write quite a bit here, I got quite a bit written on the chapter I'm
working on for the O'Reilly book. So I probably won't get much added here
today.
I did decide to add a Book
of the Week feature. I read quite a few books, typically one a day or
thereabouts. I decided that, as an experiment, I'd write a short review
each week of the best book I read the preceding week, assuming there was
one worthy of notice. I'll put a link to Amazon.com with each book. If you
want to buy the book, clicking on that link will get it for you directly,
and get me a small percentage from Amazon that I apply to the costs of
running this site. The first Book of the Week is The
Gates of Fire, by Steven Pressfield, an account of the Battle of
Thermopylae in 480 BC. I may stick with this format, or I may just start
embedding the short book reviews in my daily journal page. We'll see how
it works out.
If I have time later, and can think of something to write
about, I'll add more to today's section later. Back to work on the book.
I received the following mail from Tom Syroid of
Saskatoon, Saskatchewan:
In the interest of expediency
-- as I know you're more than a little bit busy right now -- I'll do my
best to be brief and to the point. <G> Sometimes, like you, I get
a little wordy in my enthusiasm...
I read your Day Notes every day, sometimes more than once. They are
entertaining, educating, and insightful. Not bad for loosely spun
material. Contrary to the comments I read today, I wish (at times) you
would write more. Don't you love Democracy?
I think your idea for reviewers is excellent. For years I have
beta-tested material for numerous sources, spent 100s of hours doing so,
and got no more than a copy of the program I had been living with for
the past months. Big deal? It was to me -- I often couldn't afford to
buy a copy myself as this is all just a very serious hobby for me, and
serious hobbies cost serious money. I appreciated what they did -- I
hope they appreciated what I did. Everyone won, which in this day and
age is a real bonus.
So go ahead and add me to your list if you think I can help. Be like
Microsoft and promise me a copy of the book when you're done (you should
be able to get eval copies pretty cheap I would think) and hopefully
everyone will win-win here as well.
For background: I'm a closet writer, read at least one novel a week
(sometimes two) on top of anything technical I can find and anything
technical that relates to my job as a mechanic. I had a good command of
the English language (I think). I've been around computers in one form
or another for 10 years now. I dual-boot Win98/NT, use Office
97/Outlook98 (soon to be Office 2000) for most of my work, and have a
high speed connection to the Internet (cable).
As I say, let me know if I can help. In the meantime, Cheers and keep
writing.
Thanks for the kind words. As far as tech
review, I'll certainly put your name in the hat. My guess is that, subject
to O'Reilly's approval, we'll end up somehow selecting a half-dozen or so
reviewers for each chapter. To keep the numbers manageable, we'll probably
end up using many of the same reviewers (particularly those that do
thorough jobs) for additional chapters. At the end of it all, we'll make
sure one way or another that each reviewer at least gets a copy of the
book, even if we have to buy and ship them ourselves. That probably won't
happen. O'Reilly are good folks, and I suspect they'll happily provide the
books and ship them. All of this is, of course, subject to O'Reilly
approving the scheme, which they haven't yet done. Incidentally, I envy
you your cable modem...
* * * * *
And the following mail, from Gary M. Berg:
I've got a quick question about
NT4 workstation and mirroring. I know that ideally I'd use SCSI drives
in this environment for best performance, but I'm curious as to whether
or not using Ultra DMA 33 drives is a practical option. I was once told
that using EIDE drives for mirroring was "terribly slow", but
I wonder if that was before the Ultra 33 drives. Given that I can buy
10Gb drives in the $300 range, I can just about purchase a pair of Ultra
DMA drives for the same cost as one F/W SCSI drive with similar
capacity.
Do you have any experience with mirroring Ultra DMA drives with NT4
workstation? How much effect on performance is there? The other
possibility would be to use the Promise Technology "FastTrack"
controller, which will do the mirroring itself, but I'm wondering if the
performance of NT mirroring on its own would be "good enough".
I'm a little bit confused. NTWS doesn't
implement mirroring in software. The only option, so far as I know, is
what Microsoft calls "Disk Striping", which is what other people
call RAID-0. That's not RAID at all, of course. It works by striping data
block-wise alternately between two physical drives. In fact, it's less
fault tolerant than a single drive, because a failure of either drive
results in losing the contents of both.
NTS4 includes support for Disk Striping (RAID-0), Mirroring/Duplexing
(RAID-1), and Disk Striping with Parity (RAID-5). You can supposedly
(illegally) implement the RAID-1 and RAID-5 options on NTWS4 by using the
Fault Tolerance Disk Driver (Ftdisk.sys?) from NTS4, although I've never
tried that.
If it was me, I'd go with the Promise card and the two 10 GB Ultra-DMA/33
drives. Relative to SCSI, you'll give up a little bit of performance, both
because UDMA yields only 33 MB/s relative to the the 80 MB/s of the
fastest SCSI implementations, and because SCSI supports some neat stuff
like command queuing and elevator seeking that ATA either doesn't
implement or does poorly. But overall, I think you'd be hard-pressed to
tell the difference in a single user NTWS environment.
I'll put your message and my response up on my web site. I probably only
have about 1% of Pournelle's readership, but there are quite a few very
competent people who read my stuff regularly. Perhaps one of them knows
more.
And more from Gary M. Berg:
No, I'm a little bit confused.
I didn't realize that NTWS didn't support the mirroring along with the
striping. In that case, the Promise card is likely the only choice, and
certainly has gotten good reviews.
The reviews I've seen on
Storage Review (www.storagereview.com)
have indicated that in the real world a DMA 33 drive versus an Ultra 2
SCSI drive are equal if the mechanisms are the same for all practical
purposes. They tested ATA and SCSI versions of the same drive and saw
very little difference in performance except when the data was coming
from the drive's cache memory.
The review of the FastTrack on
Storage Review is oriented towards striping, not mirroring. They
indicate that for maximum performance you should put each drive on its
own IDE channel; unfortunately that leaves no good place to put the
CD-ROM drive (it should interfere and slow down the IDE channel it's
attached to, right?).
Putting a CD-ROM drive on the same channel as
an ATA hard disk will not necessarily slow down the hard disk. In fact, if
the hard disk, the CD-ROM drive, and the ATA interface itself are all
recent, you won't have any problems at all. The problem you're referring
to occurred on older systems when the ATA interface was not capable of
using independent timing for master and slave devices. In that case, the
only solution was to set the timing at the least-common-denominator value
that the CD-ROM drive supported, thereby crippling the hard disk transfer
rate. Any modern interface (e.g. the Intel PIIX3 or PIIX4) can use
independent timing for master and slave. Since you plan to install a
bus-mastering ATA controller that will certainly support independent
timing, you won't have any problem at all.
Friday,
November 20, 1998
I got quite a bit done yesterday on the chapter I'm
working on for the O'Reilly book, but I have lots more to do. On previous
books, I've typically completed the first draft at about a chapter a week.
This one's taking more like two weeks per chapter, because the chapters
are shorter and therefore take longer to write. If that sounds stupid,
it's not really. I still write the initial draft to the original length,
and then have to go back and nip and tuck to shorten it. So I probably
won't get much added here today.
* * * * *
I notice that CNET reports that an internal Microsoft memo
presented at the trial suggests charging Windows users an annual annuity
starting in 2001. Microsoft is starting to annoy me. I'm a strong
supporter of the free market and laissez-faire Capitalism, but that
doesn't mean I always have to be happy with the results.
Microsoft is starting to act desperate, and there may be
some good reasons for it. After all, they're running out of potential
customers. Office provides about 40% of their revenue, and people are
starting to wonder why they bother to upgrade. Sub-$1,000 PCs (IBM just
introduced a $600 one) are cutting a big hole in the revenue Microsoft
gains from bundling. After all, people will eventually start balking when
the cost of Windows and Office becomes the majority of what they're paying
for a PC.
Then there's NT5, which Microsoft has bet the company on
and can't get out the door. Even if they can, it's not going to be the
product that they'd hoped, and Linux lurks on the horizon as a deadly
threat. After all, how can they compete with an operating system that's
stable, well-supported, and free? Given a directory (NDS for Linux,
perhaps?), Linux becomes a viable alternative NOS, making it a three way
fight between NetWare, Linux, and NT. Actually, it's a two way fight,
because NT5 still doesn't have a real directory. ADS is
essentially just a directory view of the ancient NT domain scheme. It's
flat, for god's sake.
So, Microsoft is right to be running scared. Gates is
smart enough to see what many people don't. Companies that rest on their
laurels often don't last long. It'll be interesting to see where Microsoft
is in five years. Chances are, they'll be just as big and dominant as
ever, if not more so. There's a possibility, however, that they'll have
shrunk to a minor player that specializes in producing Linux applications.
A lot of people would love to see that.
* * * * *
And the following mail from Roger G. Smith about the
Promise ATA RAID card:
I've been following the Raid/FastTrack
controller discussion. Your reader seems to be leaning to the Promise
FastTrack A/V controller and should get outstanding A/V performance out
of this card, especially when you consider that you can run two, three,
or four drives in RAID 0. Four UDMA drives in that configuration should
provide enough thru-put to saturate any application I can think of.
That may well be. I've never worked with A/V, although I do
know there are some special SCSI drives designed for it. UDMA drives in
RAID 0 array should indeed be able to saturate any application, though.
I'm just not sure I'd want to count on RAID 0 for storing anything other
than temp and scratch files. There's too much risk of at least one of the
drives in the array failing, which loses all the data in the array.
You did not suggest, nor I would I use this
controller for a CD-ROM drive even if it supports ATAPI CD-ROM, which I
doubt. Any PCI motherboard should have two built in EIDE channels, at
least one of which can be used at the same time as the Promise card. Put
an inexpensive hard drive for boot and miscellaneous software with a
slaved CD-ROM on interrupt 14, disable the built-in secondary EIDE
channel, and configure the Promise FastTrack controller as int 15. Or,
put the Fastrack on the primary and only the CD-ROM on the built-in
secondary channel to save a few dollars and a drive bay, but I'd just as
soon use the array for A/V streaming only, if that is my intended
application.
That's a good point, and I hadn't really mentioned it in
enough detail. ATA and ATAPI aren't the same thing, and an interface has
to explicitly support ATAPI if it is to support IDE CD-ROM drives, tape
drives, etc. Some "smart" controllers do, others don't. I don't
know about the Promise. If it does support ATAPI, though, there's really
no reason not to use a CD-ROM drive on it (other than that you'd be using
a valuable RAID-capable channel for a CD-ROM). The real issue is whether
the interface supports independent timing for masters and slaves. If so,
using a CD-ROM drive won't slow down a hard drive on the same cable.
Nearly all modern interfaces support independent master/slave timing.
As far as disabling one of the embedded ATA ports, he may
not have to. Many people aren't aware of it, but in addition to the
standard ATA interface 0 and interface 1 settings, there are semi-standard
settings for ATA interfaces 2 and 3. The third ATA interface is assigned
to IRQ 12 or 11, with channel 0 assigned base address 0x1E8 - 0x1EF and
channel 1 0x3EE - 0x3EF. The fourth ATA interface is assigned to IRQ 10 or
9, with channel 0 assigned base address 0x168 - 0x16F and channel 1 0x36E
- 0x36F.
Assuming that the Promise supports tertiary and quaternary
ATA interfaces, there's no reason he shouldn't be able to run eight
ATA/ATAPI devices on one PC. He'll obviously need either BIOS or driver
support for the additional channels. I understand that recent Phoenix
BIOSs allow ATA ports beyond 0 and 1 to be defined arbitrarily, although
I've never tried it.
On a different subject, I didn't realize
that NTWS had _any_ version of RAID whatsoever.Keep those tips, info and
observations coming.
Well, it doesn't, unless you consider "RAID-0" to
be RAID, which it ain't. I remember years ago that many SCSI host adapters
had a simple jumper setting that allowed you to mirror drives, creating a
true RAID-1. The reason that's not done with IDE nowadays has to do with
some technical constraints that limit when multiple ATA devices on one
interface can read and write, which is one of the reasons that NT software
RAID is so slow, particularly on ATA drives. A hardware RAID solution is
always better, particularly in an ATA environment. Although I don't have
any experience with the Promise card, it sounds like he'll be happy with
what it can do.
* * * * *
With Barbara quitting her day job, I'm thinking again
about moving to Vermont. Actually, not so much moving to Vermont as just
moving, period. Ideally, I'd like a place without government. Failing
that, I'd like a place where the government raised minimal taxes, provided
minimal services, and had few laws. Given my druthers, I'd move to
Heinlein's Luna or Pournelle's Sparta in a heartbeat, but I don't know how
to do that. What I want is someplace that's free, but there's nowhere like
that left on this planet. Barbara refuses to consider emigrating to
another country anyway.
What first made me think about Vermont was their gun
control laws. They don't have any. I like that in a government. Of course,
their complete lack of gun control laws means that they also have the
lowest crime rate, but try telling that to the anti-gunners. At any rate,
it's not their lack of gun control laws per se that attracts me,
but the underlying thinking. Obviously, they feel no compulsion to pass
laws to solve non-problems.
I had this confirmed last night. A banner ad really
annoyed me for some reason, and I remembered reading about the Internet
JunkBuster. I hit their site to check it out. I can't really use it
because it's a proxy server, and I already run the WinGate proxy server to
provide shared Internet access on my network. But while I was there I
found a page of interesting privacy links. One of them concerned state
laws regarding taping telephone conversations. Some states only require
that one party be aware of the taping, others requires both, and some have
special laws about wireless phones. When I got to the page, it had a
section for each state, with various icons to summarize the state laws on
the subject, and a brief text summary of relevant civil and criminal
statutes, fines, and so on. When I scrolled down to Vermont, there was
only a "1" icon, indicating that anyone could tape anything he
wanted. In the box that summarized state laws, all it said was "No
Law". I like that in a government.
So perhaps Barbara and I will eventually move to Vermont.
We both have parents here, but someday I think we'll up stakes. She liked
Vermont when we visited it ten years or so ago, and I think she'd be happy
living there. I know I would. The ten feet of annual snow fall won't
bother me in the slightest. I don't leave the house much anyway. What I'd
really like to do is buy a place that straddled Vermont, New Hampshire,
and Canada. Vermont for its small government and lack of laws. New
Hampshire for its lack of personal income tax. Canada for a place to flee
when the U.S. federal government finally becomes completely totalitarian.
* * * * *
And more from Gary M. Berg:
I believe when I was poking around on the
Promise Technology site that the FastTrack controller does support being
the third IDE channel - in fact it's Plug 'n Play and/or gets the IRQ
assigned by the PCI slot. It may or may not be bootable based on what
sort of boot options the motherboard supports.
It looks to me as though it can be the tertiary and
quaternary controller, at least going by the diagram that shows arrows
pointing to both embedded ATA interfaces on the system board and
commenting that you can continue to use them. They didn't have any really
technical documents that I could find, so I'm not clear on several issues,
not least of which is how many IRQs the board requires. It apparently
appears to the system as a first-party (bus master) DMA device and is
listed elsewhere as a single-channel RAID controller, but I'm not sure if
that means that it needs only one IRQ to support four drives. I have a
call in to their Marketing Director, whom I hope to convince to send me an
eval unit.
I also notice that the board supports several technologies
I thought were limited to SCSI. They explicitly mention elevator seeking
and scatter-gather I/O. I'll try to get my hands on one to find out the
facts.
My interest in the board is not for an A/V
application (using Raid-0) but for redundancy (Raid-1). I don't want to
lose any data to a drive failure. IRQs in machines are getting scarce:
Irq 5 - Sound board
Irq 9 - Video board
Irq 10 - Network Board
Irq 11 - SCSI board for tape and MO and
CD-RW
Irq 14/15 IDE controllers
and somewhere in there you'd like to toss
USB, even though for an NT system it's a moot point. To put the Promise
board in you'd have to disable the primary IDE controller and continue
to run a CD-ROM on the secondary IDE controller.
Yep, you're reasonably full, but that still leaves you the
legacy IRQ 3 and 4 if you can dump one or both of your serial ports, and
possibly IRQ 7, 8 and 13. It sounds like you should be able to fit it one
way or another.
Other than buying NTS4 to run on a
workstation computer for Raid, there don't appear to be any other
options using IDE drives. You can purchase an Adaptec Raid controller
for SCSI for about $500 (about the NTS4 premium!) as long as you use
SCSI drives. But, not only is the controller more expensive, but the
drives are too. I can buy a 12Gb Quantum EIDE drive for $250 while a 9Gb
IBM SCSI drive is about $500. Sigh.
No, there don't appear to be many options. I was sure that
Adaptec at one point had done an ATA RAID board (or at least announced
one). I was thinking there were one or two others as well, but I can't
locate them. This discussion has gotten me interested enough in the
Promise to do some poking around with this.
Saturday,
November 21, 1998
I ended up spending an hour on the phone yesterday with
the marketing director of Promise, and I must say that I'm impressed.
These folks seem to have a mission to do in SCSI. Their FastTrak card not
only implements RAID 0, RAID 1, and RAID 0+1, but implements quite a few
features I thought were SCSI-only, including elevator seeking, load
balancing, a hardware scatter-gather engine, and target command queuing.
For workstations and small servers, this product appears to provide about
99% of the benefits of a SCSI disk subsystem at a small fraction of the
price. I should have an eval unit on the way to me shortly.
* * * * *
And I've been bitten by the problem Pournelle described
long ago about slow response in the FrontPage Editor. I'm a touch-typist,
so it doesn't impact me as much as it might others, but there is now a
noticeable delay between the time I type a letter and when that character
appears on screen. Similarly, backspacing over a mistake causes a
noticeable pause before each character is deleted. Pournelle speculated
that using tables was the cause of the problem, and he appears to be
right. I never had any lags at all using the old page format. Only since
I've started to use tables has this problem manifested. This on a 300 MHz
Pentium II with 128 MB.
* * * * *
And the morning paper reports that an 18 year old laborer
who worked at a local warehouse wanted to leave work early. He decided
that burning the place down would provide a good excuse, so he started a
fire that ended up destroying the building and causing $34 million in
damage to the building and its contents. No one was injured, at least. The
reports said that he was a nice boy and not "disgruntled." I'd
hate to see him really mad.
We're off to the Tucker's for dinner tonight. It'll be
nice to get out of the house for a change. I've been so busy the last
month or so that I've barely left the house. I wonder what new computer
toys Steve has to play with.
Sunday,
November 22, 1998
Pournelle is back and
updating his web site. He's not going to the Hackers' conference as he'd
planned, and it's because he had a serious automobile accident on the way
back from Comdex. He rolled his Bronco II in Death Valley and ended up
having to hike 23 miles to the nearest road. He was shaken up and has some
cuts and bruises but otherwise came through the ordeal okay.
* * * * *
We had a nice time at the Tucker's house last night. Barbara took a
bunch of her Canada photographs over to scan. Although she ordered Kodak
Photo diskettes for most of the rolls she shot, she had the first couple
of rolls processed normally, and her dad had shot quite a few without
getting diskettes made either. My scanner is temporarily hors de combat,
so Steve agreed to do the scans for Barbara. A flatbed scanner turns out
not to be ideal for scanning drugstore prints. Steve had a terrible time
keeping each print from floating away before he could get the lid down to
secure it.
While they were back in Steve's office doing that, the Tucker's
six-year-old Andrew and I were watching a rental movie, the name of which
escapes me. It didn't have much of a plot, but involved action figures
coming to life and going to war with each other. At the end, the
heroine/love-interest ran over the bad-guy action figures with a riding
mower. It's no wonder the thing was rated PG. All adults were clueless, as
usual. What are we teaching these kids?
* * * * *
And today is the first on the new schedule. Barbara used to work every
Wednesday evening, and spent the morning cleaning house, which made it
hard for me to get any work done. I used to do the laundry on weekends,
splitting loads between Saturday and Sunday. Now that Barbara will be
working at home, we've decided to make Sunday the day that we'll do all
the house work. She can go back to playing her weekly round of golf with
her dad on Thursday instead of Sunday.
And I'd better get to work. The first load is drying and the second
washing, but I have some software to install on my own system and on
Barbara's. The first thing is Quicken 99. I'll install and test it on my
system first before I risk screwing up Barbara's real Quicken
installation. She's running something like V4 for Windows, and it isn't
Y2K compliant. I also need to get a copy of FrontPage 98 for her so that
she can maintain her own pages.
Coming Soon (I hope)
|