Computer reliability

05-16-2011, 12:37 AM

Quote:

Originally Posted by Philip

As systems get more and more sophisticated, unwanted side effects are more likely to creep in - IMHO.

I would only agree up to a point. I have been programming mainframe computers since 1975.

I can remember the first time a customer reported staying up for 7 full days before rebooting (actually called an IPL Initial Program Load0.

These days I believe the reliability is 5 nines. That is 99.999 percent of up-time, not counting planned outages. That is 5 minutes per year of unscheduled downtime. a very complex system with very high reliability, but no moving parts and that helps a lot. But the mainframe has been around since 1965, so a lot of knowledge has been gained over those years.

In my opinion, the PC or laptop that you are using could be made just as reliable. But you would not like the price.

My son is a mechanical engineer. He has to get it right the first time. At 22,300 miles above the earth you don't make service calls. I have seen the satellites during assembly then definitely could as complex and they do have moving parts.

I don't think it is the complexity of the system that is the issue. I think it is the budget of the buyer.

**Bill** · 05-16-2011, 09:11 AM

Wayne -

Having worked in the space industry, I have great respect for the idea and achievement of 5 nines reliability. And actually, I don't think that the hardware in todays PCs is far from that, based on my ongoing experience with the 5 machines I currently own and use.

My observation is that what crashes is the software, usually the Operating System (OS) as much as anything. Windows 3.1 was actually pretty good. Then came Windows 95 - remember when Bill Gates crashed it during the introduction ceremony in Redmond? That was terribly embarassing for him, but came to be viewed as common. My experience with Win95 was that it probably wouldn't keep going for a full 12 hours, but if it did, you had to shut it down at the end of the day, or there was inevitably a BSOD in the morning. Win98 improved this considerably, and overnight operation was not necessarily instant death.

WinXP seems to have fixed most of this. I haven't had an OS-related BSOD in a couple years, though I admit I usually shut down at the end of the day. I haven't heard that Vista or Win7 are any better, so I continue to use XP.

So your comment that "I wouldn't like the price" is quite correct - but I think the hardware is way ahead of the software in this respect. Wish it would catch up.

Bill

Mr. Adventure · 05-16-2011, 09:37 AM

Quote:

Originally Posted by Bill

Wayne -

Having worked in the space industry, I have great respect for the idea and achievement of 5 nines reliability. And actually, I don't think that the hardware in todays PCs is far from that, based on my ongoing experience with the 5 machines I currently own and use.

My observation is that what crashes is the software, usually the Operating System (OS) as much as anything. Windows 3.1 was actually pretty good. Then came Windows 95 - remember when Bill Gates crashed it during the introduction ceremony in Redmond? That was terribly embarassing for him, but came to be viewed as common. My experience with Win95 was that it probably wouldn't keep going for a full 12 hours, but if it did, you had to shut it down at the end of the day, or there was inevitably a BSOD in the morning. Win98 improved this considerably, and overnight operation was not necessarily instant death.

WinXP seems to have fixed most of this. I haven't had an OS-related BSOD in a couple years, though I admit I usually shut down at the end of the day. I haven't heard that Vista or Win7 are any better, so I continue to use XP.

So your comment that "I wouldn't like the price" is quite correct - but I think the hardware is way ahead of the software in this respect. Wish it would catch up.

Bill

There was an announced bug in Windows 95 that would make it crash every 42 days or something like that. I always wondered how they could possibly have ever kept it running long enough to find that out.

Win-7 is faster, very solid, much quicker startup and shutdown, has better memory management, and it uses all those nifty processors in the new hardware. But it won't run lots of old programs without running them under an XP emulator in the "Professional" version. Reminds me of PopBeavers' mainframes running VM, except there's no staff or training and 10,000 times as much data center.

Philip · 05-16-2011, 03:31 PM

Hello Folks,

Well I just couldn't resist entering this interesting chat about hardware/software reliability. I tend to side with Bill. Computer hardware is way ahead of software in reliability. Even the cheapest computer in todays market is very durable and quiet fast for most applications. The Microsoft operating systems(plural) are another thing altogether. I have been through them all and I am now stuck with Windows Vista which is awlful. It has no stability and continually has one new glitch after another. The best experience I had was with XP which at least did not frequently crash.

I have made my mind up that I will buy an Apple the next time around. I have quite a few friends who were Microsoft and are now Apple. None of them have had any operating system problems.

Getting back to auto control systems - what I was trying to say by my statement "As systems get more and more sophisticated, unwanted side effects are more likely to creep in" was not that the computer hardware or even software fails. The problems arise from the design of the sophisticated and very sensitive system functions. Unless every possible normal driving situation can be anticipated in the system design, sensors can tell the system something is happening when it is not actually happening - just a normal driving situation. The system then goes about trying to make corrections for something that is in fact not happening and the driver has no control over these actions. I realize that these situations are rare or non existant in a well designed system but it is very difficult for software and sensors to know all that is happening in the real world of driving. The more deeply a system intrudes into everyday driving functions, the more the system must be able to make fine distinctions between normal and abnormal operating situations.

An interesting discussion.

Take care,

Phil

**mjlaupp** · 05-16-2011, 03:37 PM

Bill,
Please move posts 12, 13, 14 & 15 to the Off Topic Forum. They serve no purpose here!

**Bill** · 05-17-2011, 07:24 AM

Agreed, Mike, and done. I had already talked with PopBeavers about it yesterday, but hadn't gotten around to actually doing it.

Bill

05-17-2011, 08:46 AM

Reliability frequently has more to do with the configurations more than anything else. Certainly the hardware today is a lot more robust than it use to be. Failures due to actual physical problems are a rariety today past the first 90 days. If it lasts that long and the environment is not hostile it is going to last for years.

What is the big killer in most systems is unstable coding. Somebody puts in some whiz-bang hardware with a weak driver in a system that has been downloading everything that is free on the Internet and expects the machine to stay rock solid...it ain't a gonna happen (or at least not for long).

If you stick to tried and true s/w and h/w, put enough memory in, a nice solid video/audio system, and some decent disks the system is probably going to do OK. My kids are not happy with me because I won't install their s/w on the computers (no admin for them). But my machines run and theirs are always having problems

That's why my W2K and XP boxes run...

Mr. Adventure · 05-17-2011, 09:20 AM

Quote:

Originally Posted by Philip

...
I have made my mind up that I will buy an Apple the next time around.
...

Win 7 is super solid, though I'd say it's "not better, just unnecessarily different" in many ways (Apple has this problem, too). And, there's the problem with the outrageous price difference (I don't mind people making money selling me things nearly so much as when they are taking 30% off the top of everything I'm allowed to use while they interfere with competing choices).

So I'll be quietly grumbling once in awhile until Android learns to use a keyboard and pointing device other than my index finger, grows a file system interface, and develops a little more substance beyond the flashy toys. Android is already well ahead along this path, and I'm hoping an open architecture will ultimately triumph over it's greedy rivals at Apple, pretty much like the last time around.

05-17-2011, 10:24 AM

I have seen systems where they verify that every line of code has been executed at least once. This is not to say that all code paths have been verified, but it does provide a level of confidence.

This is much easier to do with a large quantity of small modules instead of a small quantity of large modules.

Sadly, I have not seen nor heard of any software that is being measured that way since 1980.

Economics.

**wbmiller3** · 05-17-2011, 05:58 PM

The shuttle's onboard software is the most reliable I know of. Every problem is tracked until it is fixed or determined to be OK. If you are interested in such stuff, here is an account of how a potentially fatal problem was encountered and fixed early in the program -

http://klabs.org/DEI/Processor/shutt...ter_system.pdf

Sadly, these lessons were not carried over to the space station, which has thousands of unresolved software issues.