SSD use

securityguy

bsod wrote: »

everyone with any experience knows a reboot cures a suddenly slow computer

That's an entirely different scenario (it's also a stupid way to fix it, because it means you lose the forensic evidence to stop it happening next time.

as for the cacheing malarky, it reads like a microsoft publicity statement for a new OS, the one that's going to be the fastest Windows yet, but never is. If cacheing was so effective, an ssd wouldn't make any difference, but it does.

No operating system could operate at a sensible speed without caching disk in RAM. What do you think RAM is, if not in large part a cache of disk? If you don't think caching (the fundamental property of virtual memory and filesystems since the 1950s) makes no difference, what do you think is in the RAM in your computer? Could you point to the OS of the last forty years which doesn't use RAM almost exclusively as a cache for disk?

SSDs improve matters on mixed workloads (because it's not cached) and on write-bound workloads (because they're faster). Otherwise, if they speed the machine up, it hasn't got enough RAM in it. SSDs are about a thousand times slower than RAM for single operations. The ideal is enough RAM to hold all the programs you use and all the files you intend to access. Once all that stuff is in RAM, the best place for it to stay is there, with write-through or write-back (depending on criticality) of changes to disk. Which is why every operating system since about OS/360 does precisely that.

bsod

the average home user doesn't have the skill to diagnose it,

https://forums.moneysavingexpert.com/discussion/5568970

nor do 90% of the administrators out there, and they don't have the time to fart about, so stupid it is not, pragmatic, practical and sensible it is.

windows 3, dos, cp/m, disk caching was optional/provided by 3rd parties and memory allocation for it configurable - allocate too much to the cache, programs won't run or machine runs like treacle

what is ram for? did you forget the little matter of programs and data that had to fit into that tiny bit of 40 year old ram, in those days the caching was reversed, the disk was supplementing minimal ram

I didn't imply cache was useless, I implied that the idea you should never reboot because you will lose all that precious cached data and your system will run slower as a result, was ridiculous. Boot, load your browser, your spreadsheet, your email client, your music player, done - 7GB of cache spare and nothing useful to put in it, because the browser does everything

securityguy

"what is ram for? did you forget the little matter of programs and data"

Not at all. Remind me, where did the programs come from? Why is it generally a good idea to execute instructions from RAM rather than fetching them every time from disk? Remind me, where did the data come from? Why is it generally a good idea to store the document you're working on in RAM while you're working on it, rather than reading it character by character from disk each time you move the cursor?

RAM caches disk. If you're living in the primitive world of early microcomputer monitors which didn't have virtual memory, they distinguished between instructions, program data and "disk cache" which was a variety of bizarre methods of dealing with what you do with data once it's no longer being used. On anything remotely serious (ie, mainframe and mini operating systems since Atlas and microcomputer operating systems of any substance) data is worked on in RAM having been pulled from disk, and stays there as long as possible. Do you think Google turn their machines off nightly to make them run faster?

A machine that's been running for a while will be keeping its browser cache in RAM. The browser history which is used to colour-code links as clicked or unclicked. All your documents. Every program you have run since the machine booted. Every library. Every configuration file. Everything: after a while, the whole working set is in RAM. Luxury for 2016: at last, it all fits.

Which is faster, starting Word up from scratch, or clicking on it in the Dock/Taskbar once it's already started? The latter, obviously: because everything is in RAM already.

bsod

when computers were chunky, and serious, for the big boys, ... from punch cards, or reel to reel tape, that predates your sudden disdain for microcomputer operating systems (monitors?) that don't fit your 40 year old argument.

It should be obvious that I already know how caches work

I doubt there are many consumers refraining from rebooting in case their cache gets flushed as a result of reading some long paragraphs, the equally daft frazzled capacitor myth has longer legs.

They can load anything they want in a couple of seconds, who cares if it's cached.

AndyPix

What you say is technically correct security guy.
And for servers, and machines that just sit there peforming the same task all day then rebooting them is pointless unless to install updates..


However, give the average user 2 weeks with a machine that hasnt been rebooted and they will have 10 crashy internet explorer processes running with no windows open, possibly a few crashy java applets, CPU running at 60%+ at idle etc etc.


Obviously this is all fixable, but the average user doesnt have the knowledge or inclanation to do so.


So the best advice for the type of user that needs to ask the question, is it it best practice to start your machine fresh every day.

Stoke

And look at that, someone (with a bit more arsed) comes along and pretty much backs up my statement.... who'da thunk it.

bsod wrote: »

back to school if you think memory leaks and runaway processes have no impact on performance

everyone with any experience knows a reboot cures a suddenly slow computer, unix or windows There is no reason to leave them on overnight, that is just a silly waste of electricity.

as for the cacheing malarky, it reads like a microsoft publicity statement for a new OS, the one that's going to be the fastest Windows yet, but never is. If cacheing was so effective, an ssd wouldn't make any difference, but it does.

I don't think securityguy was saying they had no impact, I think he was saying pretty much what I said, which was that most of your (andy+neils) arguments are utter nonsense. Which they are.

To be fair, you pseudo-intellectuals who have had a go at me, but now securityguy has come along and put forward my arguments more eloquently, it's funny how now you want to debate or 'change your stance a little'.... and are you sure you even mean a runaway process? I think you mean a zombie process..... which means again you have absolutely no scooby what you're talking about do you? Without trying to academically !!!!!-slap you here, a proper runaway process will usually get stuck in an infinite loop that it usually can't break out of, at least without the help of the init process. Most good OS's can actually detect this and kill them. Ubuntu certainly can. So Does Mac, and I think Windows does as well....

You're talking about zombie processes I would imagine, which usually linger in an exit state but don't free and therefore remain a spawned process of the init..... You're right, they do have an impact on performance, but it's pretty small and really, I would be trying to work out why a process/daemon is being left in a zombie state. Perhaps a problem with systemd in linux, or a development bug in windows.

How much electricity do you actually think a modern PC uses if you leave it on overnight? Not much. You'll probably use more electricity waiting for it to boot up to a fully usable OS.

hahahahah, hilarious. Your lack of knowledge about caching is brilliant

Stoke

AndyPix wrote: »

What you say is technically correct security guy.
And for servers, and machines that just sit there peforming the same task all day then rebooting them is pointless unless to install updates..


However, give the average user 2 weeks with a machine that hasnt been rebooted and they will have 10 crashy internet explorer processes running with no windows open, possibly a few crashy java applets, CPU running at 60%+ at idle etc etc.


Obviously this is all fixable, but the average user doesnt have the knowledge or inclanation to do so.


So the best advice for the type of user that needs to ask the question, is it it best practice to start your machine fresh every day.

You really have no clue do you...? I'm trying to be kind of nice here, but it's hard when you've all been having a go at me, and now you're sort of changing your attitude a bit.

10 Crashy internet explorer processes? Why?
A few crashy java applets? Why?
CPU running 60% at idle? Why?

Why are all these things happening? You remind me of one of these bodge mechanics who prefers to use a fix-in-a-can rather than actually fix the problem.

All of what you have said is fixable, and ironically enough, pretty much all of what you have pointed to is defective software, so it's not the average users job to fix it. If Internet Explorer has crashed 10 times, then I would be wondering why (or more likely switching to a more stable browser).

If my CPU is idling at 60%, then it's technically not idling because an idle process should be handling that, but I would be trying to work out what is going wrong.....

securityguy

Stoke wrote: »

You're talking about zombie processes I would imagine, which usually linger in an exit state but don't free and therefore remain a spawned process of the init..... You're right, they do have an impact on performance, but it's pretty small and really, I would be trying to work out why a process/daemon is being left in a zombie state. Perhaps a problem with systemd in linux, or a development bug in windows.

Actually, zombies never consumed resource of any substance and don't impact performance measurably. In old-money Unix they have a slot in the process table but don't have a u-area (where things like arguments and filedescriptors live). Which is why you can't on old-money Unix see which command left the zombie, because that information isn't in the process table, but is instead in the u-area. Linux does some slightly trickier things to retain a bit more information, so you can at least see the process name even for zombies. But a zombie can't claim any memory (as it doesn't have a u area to store the mappings) and can't be scheduled onto a CPU (as it doesn't have any instructions, or a stack).

On this occasion, I suspect our bsod/Andy friends are right, and they are talking about processes which are spinning CPU in userspace. OSX doesn't kill them automatically: for example, there's a bug in OSX ntpd (or, perhaps more accurately, a bug in ntpd on OSX) which means that if you have a non-standard configuration and sleep and wake the machine on different networks, it will very occasionally get itself into a tight loop looking for changes to the interfaces. On the rare occasions this happens, I kill it, but I can imagine some people would reboot. But whether the machine has been up for an hour or a month doesn't affect the probability with which it happens, so "when the machine is spinning the CPU I reboot" is a lot more sensible than "I reboot the machine to stop things spinning the CPU, even if it isn't currently doing so".

[quote[
How much electricity do you actually think a modern PC uses if you leave it on overnight? Not much. You'll probably use more electricity waiting for it to boot up to a fully usable OS.
[/quote]

Very little to none, in my case. It goes into a sleep state after 10 minutes where all that has to happen is the RAM remains powered, and then after an hour it flushes the RAM to SSD and powers off completely. So I never reboot the machine unless it's an update or something very bad has happened, and I don't consume any power overnight anyway. For some reason, hibernation is another thing that upsets some people.

Stoke

securityguy wrote: »

Actually, zombies never consumed resource of any substance and don't impact performance measurably. In old-money Unix they have a slot in the process table but don't have a u-area (where things like arguments and filedescriptors live). Which is why you can't on old-money Unix see which command left the zombie, because that information isn't in the process table, but is instead in the u-area. Linux does some slightly trickier things to retain a bit more information, so you can at least see the process name even for zombies. But a zombie can't claim any memory (as it doesn't have a u area to store the mappings) and can't be scheduled onto a CPU (as it doesn't have any instructions, or a stack).

On this occasion, I suspect our bsod/Andy friends are right, and they are talking about processes which are spinning CPU in userspace. OSX doesn't kill them automatically: for example, there's a bug in OSX ntpd (or, perhaps more accurately, a bug in ntpd on OSX) which means that if you have a non-standard configuration and sleep and wake the machine on different networks, it will very occasionally get itself into a tight loop looking for changes to the interfaces. On the rare occasions this happens, I kill it, but I can imagine some people would reboot. But whether the machine has been up for an hour or a month doesn't affect the probability with which it happens, so "when the machine is spinning the CPU I reboot" is a lot more sensible than "I reboot the machine to stop things spinning the CPU, even if it isn't currently doing so".

[quote[
How much electricity do you actually think a modern PC uses if you leave it on overnight? Not much. You'll probably use more electricity waiting for it to boot up to a fully usable OS.

Very little to none, in my case. It goes into a sleep state after 10 minutes where all that has to happen is the RAM remains powered, and then after an hour it flushes the RAM to SSD and powers off completely. So I never reboot the machine unless it's an update or something very bad has happened, and I don't consume any power overnight anyway. For some reason, hibernation is another thing that upsets some people.[/QUOTE]

Not sure I agree with that. The entire process may still be loaded into the RAM, stack or not really isn't important. The process could still be in the RAM, even when it's in an exit state. I think if it's in a defunct state then no RAM is being used.

If they're talking about processes spinning in the userspace (a runaway process), then it depends. Being in a tight loop doesn't necessarily cause a huge dent in performance. It's going to depend on a number of things. That said, a runaway process can also cause a huge amount of CPU usage, so I begin to wonder at what situation they're talking about.

AndyPix

Stoke wrote: »

And look at that, someone (with a bit more arsed) comes along and pretty much backs up my statement.... who'da thunk it.

Security guy did nothing of the sort, he made completely different argements to yourself and mentioned nothing of component expansion and contraction, which was the only point you made, even when pressed.

Stoke wrote: »

I don't think securityguy was saying they had no impact, I think he was saying pretty much what I said, which was that most
of your (andy+neils) arguments are utter nonsense. Which they are.

No, they are not. I have worked in this field for 20+ years and still do, and what i have said comes from actual physical experience in real world scenarios that i see on a daily basis.
Not from an ancient phd in computer science , learnt from dusty out of date text books that is meaningless really unless you want to go into acedemia.
The theory is sound, but doesnt translate to modern or real world situations.

Stoke wrote: »

you pseudo-intellectuals who have had a go at me

Im not claiming to be an intellectual, far from it. I am however claiming to be an expert in real world pc issues in 2016. And no body had a go at you, they simply asked you to back up your rather rude comment with a few facts, which you couldnt do.

Stoke wrote: »

put forward my arguments more eloquently

Security guy made a completely different point to you and made no mention what so ever of component fatigue

Stoke wrote: »

it's funny how now you want to debate or 'change your stance a little'

Im happy to debate, if you actually have something to debate, my stance however remains firm

Stoke wrote: »

and are you sure you even mean a runaway process? I think you mean a zombie process

And i think you have been googling buzzwords so you can pretend to have some useful input, what the errant process is called can be anything you like. (a broken process ??) the technical details stay the same .. This comment just further emphasizes the fact that you have no idea what you are talking about and is frankly childish

Stoke wrote: »

a proper runaway process will usually get
stuck in an infinite loop that it usually can't break out of, at least without the help of the init process

This comment made me chuckle, so thanks for that - It basically translates to :-
"I have been frantically reading google to try to find something that sounds intelligent to say in this debate - and i have come up with this"

Stoke wrote: »

How much electricity do you actually think a modern PC uses if you leave it on overnight? Not much. You'll probably use
more electricity waiting for it to boot up to a fully usable OS.

Now this is just a guess, but my IT spider senses tell me that the OP was totally not thinking about electricity use when they asked their original question, and was more talking about "what is best for my computer to keep running smoothly" (please correct me if im wrong OP).


And as i said above, for the user who needs to ask this question, the best advice is for them to start their computer fresh every day. For the reasons i have given above.


Im not looking to argue and iv said all i will say on this thread now, readers can take from it what they will


Andy


Edit : SecurityGuy : no disrespect to your PHD, that's some achievement. But im sure you will agree that this is made up mostly of theory. And in theory yes, there is no reason to be restarting a computer atall apart from updates or other installs.
But in the real world, things are never like this.
If you need proof of this, visit one of your "less IT savvy" friends house, and have a look at task manager on their PC - Then come back and tell me that PC wouldnt benefit from a re-boot !!