OT: Yet Another Unhappy Customer for Vista

Return spring? Do they have springs? The servo would have to fight it all the time.

John

No springs. The head-arm assembly is balanced in three dimensions.

--
  Keith

Windows being used as an RTOS is very much like a stock yugo with 95K miles running on two cylinders being used as an F1 racer or the world renowned chief surgeon of boston general, who never finished high school.

nonexistant.

Yes, but a correctly designed journaling file system doesn't need to check every damn cluster when incorrectly shut down.

My Office PC has some hardware problem causing it to seize up occasionally, independently of the OS being run, and I have to do a hard reboot. On bad days this could happen every hour or so which is why I now set some values in the "Performance" to ridiculously conservative levels.

Anyway, I'm mostly running Linux on this thing with ext3 filesystems, and they always come up without a hitch, requiring only a few seconds of "journal recovery" on startup.

Of course in this scenario the disks themselves always have enough time to flush their internal buffers, so at least their content is consistent with the operation of the journaling FS.

If this happens during Windows sessions, it's sometimes time for an extra cup of coffee... but usually not bad; this is W2000/NTFS. But in the low-performance mode, you can really hear SolidWorks' gears grind...

--Daniel

That's because DOS really isn't an operating system. It's just a loader for one single application, and after loading only that application is responsible for anything that is going on in the entire machine (with the exception of a few tiny TSRs). And if such an application is written in a crash-proof way (which is easy), nothing can go wrong.

With modern multitasking OSes there is no clear-cut division between RAM and hard disk any more. Frequently used parts of the disks are actually kept in RAM, and infrequently used chunks of RAM get deported to the disk. It is this mixture that gives big performance benefits, but it is also this mixture that needs to be untangled on shutdown. And if important configuration data is affected, this may bring the whole system down for good.

There's no need for this kind of OS in a dedicated computer which runs a single application for a specific purpose.

Unix systems, by the way, have a strict policy to not allow applications to mess with their system-wide configuration files, all of which are ASCII text files and are kept in a dedicated directory tree. Therefore it is virtually impossible to bring down such a system by file system corruption, unless of course the power outage happens while someone is editing such a file. The Windows registry is a whole different can of worms...

robert

They don't have springs (says someone who likes to open up dead disks and marvel at the engineering inside). Maybe they take the power required for parking from a big internal cap, or they use the spinning platters as generators.

robert

Learn to spell, child.

Do you consider an embedded OS to be an RTOS?

You and reasoned logic = non-existent.

That's the whole idea behind the journal, dufus.

On a sunny day (Tue, 05 Jun 2007 21:22:56 GMT) it happened Joerg wrote in :

A running OS (and some software) has data in buffers in RAM, it uses 'caching', and if you simply plonk the power switch that data is not yet written to disk (non-volatile memory) and you loose any or all of what you are doing (caches can be quite big these days).

sync sync sync (for the very old Unix fans).

Vart Dader

I remember reading about early IBM 2.5" notebook drives which were some of the first to unload the heads (park them lifted off the disc surface).

They said when power was unexpectedly removed power from regenerative braking of the disc was used to unload the heads. This was called an 'emergency unload' and the drives were specified for a relatively small number of emergency unload cycles.

The main advantage of unloading the heads was avoiding them sticking to the disc surface preventing the disc spinning up.

I bet it is at Google... they have a lot of smart people there, and they're supposedly running something in the ballpark of 100,000 servers worldwide. At that level, I bet it's *much* cheaper to pay for a small team of full-time employees to run around changing out failed servers than it is to pay the

*significant* premium (often 5x or more compared to what I suspect Google gets) to purchase "server grade" equipment that's probably really not nearly 5x as reliable. as the cheap stuff.

For most operations I think this is a decent strategy. The main hitch that comes in is that computer technology changes quickly enough that -- unlike, say, fleet cars -- it's unlikely you're going to be purchasing the same, exact PC configuration six months from now that you would today. The bigger guys like Dell and HP keep the same motherboard/case/etc. around for a couple of years with their business-oriented PCs, but the processors they drop into them get faster... and of course they'll try to get you to "upgrade" with Vista, but at least are smart enough to offer XP when enough people scream. :-)

(I also imagine that, in John's case, he's looking at paying no more than a 2x premium for a "good" PC compared to what the average office user has...)

Which version are you running? It sure seemed to me that MS Office 97 or 2000 were about the best; XP (2002?) and 2003 seemed significantly flakier... although supposedly 2007 might be better again.

I do remember reading about the Apple Lisa in BYTE magazine a million years ago and -- at the time -- thinking how odd it was that powerdown was a software-initiated action and how "just throw the switch!" was no longer considered kosher. Still, at the time I was sitting around using Apple II's, and one of the first things anyone told you about them was to never hit the power switch when the floppy was being accessed, of course. In a modern OS the problem is that you have no way of *knowing* when the OS is going to "hit the disk" -- too many processes running in the background!

(I also remember the first time I encountered a UNIX PC that I personally could power on and off... it was a nasty old IBM RT PC running AIX, and it took so long to boot due to file system checking that I was almost certain the first time I booted it that it was broken!)

---Joel

I never wanted to use it as an RTOS. But still I did expect better than a Yugo ;-)

--
Regards, Joerg

http://www.analogconsultants.com

If you run SolidWorks on this thing I'd seriously consider checking the RAM in there (cold spray), the RAM sockets, other connectors, and if that doesn't do the trick at least buy a new motherboard. Or you could attach one of those bumper stickers from the US: "Honk if something falls off" ;-)

--
Regards, Joerg

http://www.analogconsultants.com

Yabbut, in the early 90's I used a DOS multitasker that let me switch on the fly between OrCad, Word and other stuff. I just had to have enough memory installed for that but in the good old days 5MB sufficed even for CAD. Now it seems with Vista you need 256MB just to write "Hello World". Want to sketch a Happy Birthday banner? Oh, that'll

BTW "multi-tasking" on "modern" OS isn't much spiffier than that DOS multitasker was. Whenever I want to continue writing an email while some dreadfully slow pdf loading occurs it doesn't let me. So what good does it really do? Ok, in the DOS days I wouldn't hear a "you've got mail" ping while drawing a schematic but oh well...

--
Regards, Joerg

http://www.analogconsultants.com

Yep, 2000. But always with the "past 97 features" disabled because some recipients seem to consider that the last stable version.

IMHO software initiated powerdown mandates are a clear case where technology took a wrong turn. I just won't design stuff like that.

--
Regards, Joerg

http://www.analogconsultants.com

And at 2x the above calculation manhours versus asset costs may become seriously lopsided. It's amazing how fast money flies out the window the minute someone has to pick up screwdriver and plies. Been there, we were cross charged for IT support so I know what it can cost. Which motivated us to buy top notch quality and not no-name gear.

--
Regards, Joerg

http://www.analogconsultants.com

It's done all the time in aeronautics. One scheme is several write plus majority decision upon read. This means some overhead but on a large jetliner on final approach into a busy place like LAX, with a few hundred people in there, a lengthy reboot after a tripped breaker is just not an option.

Yep, that's another method.

--
Regards, Joerg

http://www.analogconsultants.com