ReactOS recovers from a low memory situation...

Here you can discuss ReactOS related topics.

Moderator: Moderator Team

Post Reply
User avatar
dizt3mp3r
Posts: 1874
Joined: Mon Jun 14, 2010 5:54 pm

ReactOS recovers from a low memory situation...

Post by dizt3mp3r »

Just a thought. Here's hoping that ReactOS recovers more gracefully from low or out of memory conditions than Windcows in general.

I was just running Yahoo widgets and an associated editor, the VB6 IDE, thunderbird, palemoon and firefox as well as all the usual background utilities on Win7 in 4gb RAM and of course it was running in excess of 95% RAM usage and may well have peaked beyond the 100 and then Windows started to swap. I don't let it happen normally but this puny machine has a max of 4gb RAM and I'm currently doing a lot. The other machines are each doing something.

However, even when all applications are closed and memory is freed up, Windows always seems to fail to fully recover from this type of event running slowly from this point onward - and the best solution is a quick reboot which restores full performance. This has always been the way in my experience, you may know better or otherwise but this is how it has always acted after Windows has started to swap.

So back to my original thought, will ReactOS recover from such a low memory situation better than Windcows NT5 or 6?
Skillset: VMS,DOS,Windows Sysadmin from 1985, fault-tolerance, VaxCluster, Alpha,Sparc. DCL,QB,VBDOS- VB6,.NET, PHP,NODE.JS, Graphic Design, Project Manager, CMS, Quad Electronics. classic cars & m'bikes. Artist in water & oils. Historian.
Quim
Posts: 257
Joined: Wed Jul 04, 2018 11:45 pm

Re: ReactOS recovers from a low memory situation...

Post by Quim »

dizt3mp3r wrote: Thu May 02, 2019 7:09 pm Just a thought. Here's hoping that ReactOS recovers more gracefully from low or out of memory conditions than Windcows in general.

I was just running Yahoo widgets and an associated editor, the VB6 IDE, thunderbird, palemoon and firefox as well as all the usual background utilities on Win7 in 4gb RAM and of course it was running in excess of 95% RAM usage and may well have peaked beyond the 100 and then Windows started to swap. I don't let it happen normally but this puny machine has a max of 4gb RAM and I'm currently doing a lot. The other machines are each doing something.

However, even when all applications are closed and memory is freed up, Windows always seems to fail to fully recover from this type of event running slowly from this point onward - and the best solution is a quick reboot which restores full performance. This has always been the way in my experience, you may know better or otherwise but this is how it has always acted after Windows has started to swap.

So back to my original thought, will ReactOS recover from such a low memory situation better than Windcows NT5 or 6?
Well ReactOS is still in alpha stage, and I don´t know if it is enough to try to test if it is better than Windows XP + recovering from low or out of memory.
We could also test anyway to see what happens (using latest oficial release 0.4.11)....
User avatar
dizt3mp3r
Posts: 1874
Joined: Mon Jun 14, 2010 5:54 pm

Re: ReactOS recovers from a low memory situation...

Post by dizt3mp3r »

I didn't really mean now. :)

What I meant was, is there an intention to replicate Windows and its limitations in this respect or do the devs see an opening where the recovery from an out of memory situation could be handled by better programming at the kernel level, improved handling of virtual memory.
Skillset: VMS,DOS,Windows Sysadmin from 1985, fault-tolerance, VaxCluster, Alpha,Sparc. DCL,QB,VBDOS- VB6,.NET, PHP,NODE.JS, Graphic Design, Project Manager, CMS, Quad Electronics. classic cars & m'bikes. Artist in water & oils. Historian.
bobbykennedy
Posts: 18
Joined: Mon Apr 09, 2018 10:21 pm

Re: ReactOS recovers from a low memory situation...

Post by bobbykennedy »

Sure, but you'd have to test thoroughly and prove that it wouldn't break compatibility with any Windows programs and drivers.
asd1!
Posts: 3
Joined: Mon Jul 16, 2018 4:20 pm

Re: ReactOS recovers from a low memory situation...

Post by asd1! »

Sounds like the ticket with the most votes on jira: https://jira.reactos.org/browse/CORE-6699
Quim
Posts: 257
Joined: Wed Jul 04, 2018 11:45 pm

Re: ReactOS recovers from a low memory situation...

Post by Quim »

asd1! wrote: Sun May 05, 2019 12:57 am Sounds like the ticket with the most votes on jira: https://jira.reactos.org/browse/CORE-6699
So the answer to @dizt3mp3r question is no. Is not implemented yet. :cry:
ThFabba
Developer
Posts: 293
Joined: Sun Jul 11, 2010 11:39 am

Re: ReactOS recovers from a low memory situation...

Post by ThFabba »

By default, we want to replicate Windows behavior.
Now if you can actually reproduce the issue reliably and determine its cause, and we consider that a bug, we may fix it. But the implications of such a fix have to be carefully considered (Microsoft's kernel developers aren't idiots, they often have to make conscious trade-offs) to ensure we're not breaking something else (potentially more important) by making the change.

Practically speaking, I would guess issues like these fall in one of three categories:
  1. A trade-off in the OS against other concerns that was taken intentionally, and cannot be fixed without breaking a different, more important scenario
  2. A valid issue whose fix requires significant refactoring and may break compatibility (and therefore has either been addressed only in a new major Windows release, or is not addressed by Windows at all [yet])
  3. A bug in a third-party driver or service, which should be addressed in that component
User avatar
dizt3mp3r
Posts: 1874
Joined: Mon Jun 14, 2010 5:54 pm

Re: ReactOS recovers from a low memory situation...

Post by dizt3mp3r »

In the VMS o/s, some of its architecture WNT copies, the concept of the working set was matched by values for WS_MAX and WS_EXTENT that allow working sets to extend to a finite max for all processes/users. I suspect that it is the lack of these sorts of controls in Windows that allows memory tracking functions to run 'inefficiently' after swapping starts. I am guessing that some paged or non-paged pool becomes exhausted/fragmented and the situation is not fully cleared (or a pointer/structure that tracks memory usage loses context) after the resource is freed.

Reproducing? I have always imagined that anyone can experience it running any version of Windows from XP to Win7 on a core2duo or below, just set aside/install 4gb or less and run more and more applications until the o/s has to page/swap current memory to the pagefile. Then free up the memory by killing the applications and observe windows running hesitantly until you do a reboot, when performance is magically restored. Should be easily observable in a VM too.

I don't think it is related to a driver issue unless it is a part of core windows that has always been present on all versions. I have not encountered the problem on Win10 but that may be due to the fact that all the Win10 machines I have ever used have had vast amounts of RAM.

It feels like a bug to me as it feels like the handling of a situation, one that it is expected should never occur in Windows as you ought to ALWAYS have more memory than you need, PC RAM being cheap and 64bit machines capable of addressing plenty of it. Being from a VMS background where being out of RAM is a normal situation with a VAX 4300 with 128 x VT200 users on a mere 256mb, reaching the upper memory limits was a daily situation that the o/s was designed to address and handle without error.

Not having upper limits to memory locations seems to be simply sloppy in comparison but the two o/s are different in what they provide, with that in mind I find it sensible to have gauges on my desktop that indicate current RAM usage to avoid this situation ever occurring.

[ external image ]

It is essential when you are running on a puny laptop such as this, a 64bit o/s on a machine that can only handle 4gb RAM. I intend to buy a new machine but this machine is so well-configured in all other respects, compact, good screen, great keyboard, fast SSD and plenty of old/new ports, solid construction that it will run until it dies and that might be years yet.

The issue is that I am hoping more attention is paid to memory management in the finished kernel, was it a trade off by the Windows kernel devs? well, it feels as if Windows has a shoddy implementation of virtual memory and so it feels like a bug to me.
Skillset: VMS,DOS,Windows Sysadmin from 1985, fault-tolerance, VaxCluster, Alpha,Sparc. DCL,QB,VBDOS- VB6,.NET, PHP,NODE.JS, Graphic Design, Project Manager, CMS, Quad Electronics. classic cars & m'bikes. Artist in water & oils. Historian.
Reactions
Posts: 211
Joined: Mon Dec 31, 2018 3:30 am

Re: ReactOS recovers from a low memory situation...

Post by Reactions »

During my testing, it seems as though the task manager refused to update the data even though it was responsive. Is this a bug? I have no idea how to test this issue if the system can't give a reliable memory readout.
Please support/share my Patreon! Helps so much! https://www.patreon.com/reactions
User avatar
dizt3mp3r
Posts: 1874
Joined: Mon Jun 14, 2010 5:54 pm

Re: ReactOS recovers from a low memory situation...

Post by dizt3mp3r »

I'd like to add this to my 50 or so reasons why users should run ReactOS or why ReactOS will eventually be better than windows..
Skillset: VMS,DOS,Windows Sysadmin from 1985, fault-tolerance, VaxCluster, Alpha,Sparc. DCL,QB,VBDOS- VB6,.NET, PHP,NODE.JS, Graphic Design, Project Manager, CMS, Quad Electronics. classic cars & m'bikes. Artist in water & oils. Historian.
Reactions
Posts: 211
Joined: Mon Dec 31, 2018 3:30 am

Re: ReactOS recovers from a low memory situation...

Post by Reactions »

dizt3mp3r wrote: Wed May 08, 2019 2:10 pm I'd like to add this to my 50 or so reasons why users should run ReactOS or why ReactOS will eventually be better than windows..
Ummm I would not do that yet, as I feel that more testing is needed.
Please support/share my Patreon! Helps so much! https://www.patreon.com/reactions
Post Reply

Who is online

Users browsing this forum: DotBot [Crawler] and 47 guests