Startup, Shutdown, Processes and Myths

[jdhurst]

There is a lot of focus in here on processes. On an NT machine, processes are "terminate and stay resident" type of objects, that, if properly programmed as services, can also start and run without loggin on. Nice feature that occurs in lots of processes.

I maintain that these processes are benign once started, consume memory (which is static and which is of no consequence if you have enough), and otherwise cause no problems in my experience. I have 80 to 85 processes running at any one time and at any one time, my CPU load is usually under two percent. I think it is a myth that the *number* of processes affects machine performance. Except at startup and shutdown, I have not seen it happen as a regular occurrence.

So what of Startup and Shutdown?

My T41 has a very fast hard drive, a 1.8GHz Dothan CPU and 768Mb of memory. Over and above all the standard Windows objects, it loads and runs at Startup:

1. All IBM Preload software except for System Update and the R&R tools.
2. Symantec Client Security Firewall and AntiVirus (instead of, and replacing Norton Retail AntiVirus).
3. Juniper Netscreen VPN
4. Access Connection 4.41 set with only two *active* profiles (any wired and my home wireless).
5. FreeMem Pro (to keep dumping out of memory programs that stay there (Office, Outlook, Explorer, QuickBooks and such like. These may have processes but I am talking of the applications here)).
6. CacheSentry Pro to keep IE cache at 20Mb. This speeds up startup quite a bit, especialy compared to a half gigabyte or so of cache.
7. iPass Remote Access services.
8. And others.
9. I disable fast user switching and use the more secure traditional Windows Login.

Here is the startup time (which is highly representative for me).
Power On -> Power On password: 5 Seconds
Password -> Windows Login Screen: 45 seconds. This includes a time of black screen that lasts for about 5 seconds
Windows Login -> Machine is running: 2 Minutes.
Machine is running -> No hard drive activity and zero CPU: 30 more Seconds.

So grossing each major step up to the nearest minute, from the time I press power on to the time the machine is stable and ready for hard use is 4 minute or less. At that point, all but three processes are running at zero CPU and ALL IBM preload software is running at zero CPU. The three processes are System Idle (meaningless), TaskManager (2 percent), Windows Explorer (1 percent).

4 Minutes is hardly super speedy, and I have never contended such in this forum. But given that the machine will run from that point for 8 to 16 hours, 4 minutes is essentially inconsequential in my work life.

So now to Shutdown:
Most of the time (95 percent of the time), my T41 shuts down reliably in about 30 seconds.

So, while your mileage may vary, and while you may look at 4 minutes as agonizing death, I see two myths:

Myth 1. The number of processes is a problem.
Myth 2. The IBM set of drivers and applications is bloatware crap.

... JDH

[Kyocera]

did I hear someone say, make this a sticky????? I vote yes

[K0LO]

I beg to disagree. Slow startup time is the issue for a lot of us. You are correct that having a large number of processes running has minimal effect on a running XP OS, but it will definitely have an effect on the amount of time it takes to get going.

I'm glad that you are satisfied with waiting 4 minutes for your PC to start up and come to an idle state. I wouldn't be. I think that the problem here is that the PC industry has been going backwards on this issue. My first encounters with PCs were with embedded Z80 processors that would start up in milliseconds. Later I discovered disk-based CP/M systems that would take maybe 4 or 5 seconds to start up. As things have "progressed" we now seem to accept PCs that take minutes to start up.

If my car took 4 minutes to start up I would not be happy, nor would I be happy if my TV took 4 minutes to start up. So what's wrong with this picture?

Why can't PCs start up instantly? OK, maybe set a design goal of 1 second maximum. This "booting" nonsense is way out of hand because nobody is paying any attention to it as a reasonable design parameter for computing hardware.

Meanwhile, some of us (maybe we're too impatient) are stuck with tweaking our systems so that they don't take so danged long to start up. Here are my two current examples:

1. X41T tablet PC -- as delivered by IBM; 5 minutes to boot up. After tweaking, 1 minute to boot up.

2. Fairly modern desktop PC with fast hard drive -- 30 seconds to boot up.

I consider both of these too slow and would like to see times more on the order of a second or so. But I definitely would be unwilling to settle for 4 minutes.

[rkawakami]

Use hibernation .

99% of the time I simply Fn+F12 my system. At home it's docked using an ethernet connection. I hibernate, take the system to work, wake-up it and then it's using my wireless connection. At night, the reverse. System wakes-up in seconds after I supply my hard drive password.

The other 1%? That's when Microsoft deems it necessary to shut down my system and re-boot.

[K0LO]

Hi, Ray:

I knew that I should have put that in my response but didn't want to get too far off-topic. Yes; to avoid the long startup times I almost never shut down my PCs. Going in and/or out of standby takes 5 seconds; hibernation takes 15 seconds. But this is still not instant-on like my PDA and 99% of the other electronic devices on the market.

[jdhurst]

<snip>
Why can't PCs start up instantly? OK, maybe set a design goal of 1 second maximum. This "booting" nonsense is way out of hand because nobody is paying any attention to it as a reasonable design parameter for computing hardware.

It would be nice if this happened, but I don't think it will any time soon. For example, if I have to restart my cable modem (happens if cable goes out), it takes about a minute to start. A roughly similar set of occurences happens (network connectivity) when Access Connections finds a network for me. So there are just so many things going, it just takes time.

A CP/M system may start fast (I had one), but I would rather wait 4 minutes for the rich set of services I have now that CP/M never had (multiple simutaneous machines as one example).

So I do not disagree in principle, and I am not trying to be argumentative, but a 1 second startup goal is simply unachievable for a modern machine with the automation most people want.
... JDH

[rkawakami]

In order to get PDA-like response times, the whole memory system of the computer would have to be changed over to Flash (or whatever takes its place in a couple of years). Relying on regular DRAM memory to keep the operating system and applications "alive" is not practical. The momentary loss of power will mean that your data is gone. If you think 4 minute boot times is bad, that would not sit well with the computing public.

You need non-volatile memory in order to do what you propose. At the moment, hibernation is about the only thing that is available.

Sorry if I've strayed slightly off topic but as your point was about startup/shutdown times, I thought this would be meaningful.

[tomh009]

And modern PDAs rarely start up instantly, either -- the days of Palm V are long gone. Do a cold boot on a Windows Mobile 5 device or a Blackberry, and you will see bootup times on the order of 30-60 seconds. My PVR takes 10-15 seconds to be ready. My HP DEC takes probably closer to 30. My HP Digital Sender is surely over a minute. My GPS navigation system is close to 30 seconds, excluding the need to find satellites.

So on my X31, I go to standby a few times each day, and reboot maybe once a week. And I have no issues there ...

[jdhurst]

I welcome all thoughts about all devices in this thread. Just for clarification, in my original post, I believe I distinguished between startup and running. I fully accept that all the processes I *want* to run takes time, and I accept the time. My point, less well made, is that once running, there is no penalty in performance.

BTW, to stray also a bit, I just shutdown my Blackberry (a network-connected PDA and phone). It took just short of 1 minute to start and stabilize with an active network connection.

It takes about 30 seconds to restart my iPAQ.
... JDH

[Kyocera]

JD put's things in a perspective......... that gives people who "expect" more that what they are getting at an exact moment in time......something to considier before they come asking "is my system broken or questions about bloatware, etc.

[K0LO]

All of the examples posted about cable modems, PDAs, Blackberrys, etc. taking a long time to start up only serve to illustrate what I'm trying to point out.

Our industry is regressing!!!

In the vacuum tube days we accepted that it took 30 - 60 seconds for the tubes to warm up. When solid-state devices became popular we had electronic devices that were instant-on. Now that the microprocessor has become ubiquitous we are back to waiting 30 - 60 seconds for the stupid things to "boot up".

Ray was on-target with his comments. It will take a different design philosophy to achieve instant-on with processor-based devices. But it can be done if the designers put their minds to it and want to make it happen. However, if all of us just accept these delays as "oh well; that's just the way it is" then things will never change.

[jdhurst]

<snip>
Our industry is regressing!!!

In the vacuum tube days we accepted that it took 30 - 60 seconds for the tubes to warm up. When solid-state devices became popular we had electronic devices that were instant-on. Now that the microprocessor has become ubiquitous we are back to waiting 30 - 60 seconds for the stupid things to "boot up".
<snip>

I take your point, but your world is different than mine.

1. It takes at least 10 minutes for my Digital Counter to be considered to be a full digital accuracy.
2. It takes 5 minutes for either my 7704a or 7904 oscilliscopes to provide repeatable accurate voltage and timing measurements
3. My vacuum tube 606B signal generator takes about 30 seconds to start and about 30 minutes to stabilize (useable in 3 minutes).
4. The A/C in either car takes at least a minute to pump out nice cold air and up to 15 minutes to cool the car.
5. Either car needs 10 minutes of driving before being modestly warm when started at -20C
6. My scanner (which does *one* thing) takes 30 seconds to be ready if the lamp is off.
7. When I start my car, no matter how long I wait, it will not go off in two different directions at once. My PC *can* do that.

Each of the things above does different things, and as in your own list, not directly comparable to a modern, multitasking PC. Still, in my world, just about nothing except for an incandescent light bulb starts up in 1 second. I don't find it regressive, but I respect that you do.
... JDH

[ryengineer]

But it can be done if the designers put their minds to it and want to make it happen. However, if all of us just accept these delays as "oh well; that's just the way it is" then things will never change.

I think we`re yet to achieve that level. Warm-up and delay time are part and parcel of these chips.

[rkawakami]

1. It takes at least 10 minutes for my Digital Counter to be considered to be a full digital accuracy.

What, no TCO?

4. The A/C in either car takes at least a minute to pump out nice cold air and up to 15 minutes to cool the car.

You actually need A/C up there? (and use it???)

5. Either car needs 10 minutes of driving before being modestly warm when started at -20C

See previous statement...

6. My scanner (which does *one* thing) takes 30 seconds to be ready if the lamp is off.

Yours is faster. My Epson takes a couple of minutes.

7. When I start my car, no matter how long I wait, it will not go off in two different directions at once. My PC *can* do that.

Depending upon how much salt is used on the winter roads, a car CAN separate into two different pieces. .

Sorry, but I couldn't resist...

Yes, I have an old Palm Zire 71 PDA which turns on instantly (when I've remembered to charge it). My old Motorola, Qualcomm and Ericsson cell phones basically were ready to make/receive calls about 5-10 seconds after pushing the power button. My new Sony/Ericsson W810i phone now takes around 30 seconds before it locks onto a cell tower. My MacBook (OSX 10.4.8 ) takes about 23 seconds before the desktop is ready. Shutdown is 8 seconds. These things didn't exist 15 years ago but now we can't live without them. The computers I've used have always taken some time to start up (ever bootstrap a Data General Nova 3?). From a lowly 80286 AST to the 1.83Ghz Core 2 Duo MacBook. Such is the nature of computing. Drastic changes in the architecture will be needed before we ever get to the "instant-on" computer. Putting the OS into firmware is a start (none of this rotating disks of glass or metal). Flash is really too slow to be useful as main memory in a computer, so newer developments in the memory field will be required; some of it may be at the molecular level. Also, software changes may be needed as well. It's always been a given that each new release of Windows requires almost the leading edge processor and a large amount of memory (RAM and disk) to be recognized as a noticable improvement over the previous generation. Code has got to get more streamlined.

The industry is charging ahead. Some of the technology to get to this instant-on system has yet to be invented or reduced in cost to be acceptable. But compared to what we had before, today's equipment is vastly superior for the cost. Remember, the first Thinkpad laptop was released about 15 years ago and cost, what, about $4,000? Today you can get FOUR moderately-equipped T61s for that price. Waiting five or even 10 times the amount of time to load DOS vs. Vista seems to be a good enough trade-off for me.

[K0LO]

As an electronics design engineer I know that faster startup is an achievable goal, though it takes vision to make faster startup a design objective, and it takes many users sending a message to the industry that we want instant-on products. But maybe I'm the only one here who thinks that is important. I still cringe, however, when I see the newest cell phones take up to a minute to boot up. Some day we will read a tragic story in the paper about a cellphone user who had a true emergency and needed to make a call right now during a life-threatening emergency. I'd hate to be the company who designed the phone that made the user wait 60 seconds extra to make that emergency call.

The electronics industry responded admirably to calls a few years ago for better battery life in portable products. Remember the early cell phones whose batteries would only last 8 hours in standby? Today with a battery of 1/4 the volume, having standby times of 40+ hours is not unusual. Why? Longer battery life as a design goal required circuit designers to get creative and come up with ways to reduce power consumption. In current mobile microprocessors the portions of the IC that are not in use are shut down to save power, clock rates are reduced when idle, and better processes produce MOSFETs with lower leakage currents.

If we can apply some of this creative energy to the goal of eliminating the mind-numbingly slow PC boot process, then we'll have something.

Great discussion, guys. I've enjoyed your comments.

[jdhurst]

My comment about the cell phone was to relate total startup times across differing systems.

In the case of my cell phone (old Motorola and now new Blackberry), I leave them on 24x7 (even while charging during every other evening). Having fast access to a phone is a necessary function as you point out. ... JDH

[jdhurst]

<snip>
You actually need A/C up there? (and use it???) <snip>

It keeps me used to the cold weather ... JDH

[k2jsv]

The world has become a very entrenched in the "instant gratification" philosophy that we see in this day and age. 20-25 years ago 2 minute and longer boot times where acceptable because the world was not in the rush that everyone is today. It's a wireless and instant on type world. All this is unacceptable because society is flawed (in this respect), not because the equipment is flawed.

I am happy when my A31p boots and is ready to go in a minute or so. Especially when I am at work, while the computer is booting and doing it's thing I am punching in, getting a cup of coffee and greeting the people that I work with. By the time that is done, my laptop is ready for work.

Fast or slow boot times... it's all about perspective. When it comes to your cell phones and other frivolities that we all have... improvise, adapt and overcome.

[WarMachine]

The number of processes is a problem.

Not the number of processes, but the processes themselves.

You can have security issues if you let unused services running.

I agree with you, when you say that 4 minutes are nothing compared to a day. But it's very long.

FYI, I use too the Windows 2000 Logon, and I must type a password.

On every machine I own, boot time doesn't exceed the minute, even on my old TP600 I had, with 192 MB RAM and a PII 333...

Have you tried to optimize you boot time by stopping services, and make a clean startup list ?

Sure you'll see a difference.



W.

[jdhurst]

<snip>
Not the number of processes, but the processes themselves.
<snip>

Have you tried to optimize you boot time by stopping services, and make a clean startup list ?

Sure you'll see a difference.



W.

I know what the processes are on startup and I *want* them there. My overall point on processes was focussed on run-time more than startup. The 4 minutes doesn't bother me. It is less than a bathroom break in a full day. However, once running, the processes in my machine (all 78 out of 80) are running at zero and cause zero problems and have zero impact on performance. Then, too, it would be impossible (yes I used the correct word) to start everything I want started in under a minute. Simple beyond the realm of reality.

... JDH