Slashdot Mirror
Why Use Virtual Memory In Modern Systems?
Cyberhwk writes "I have a system with Windows Vista Ultimate (64-bit) installed on it, and it has 4GB of RAM. However when I've been watching system performance, my system seems to divide the work between the physical RAM and the virtual memory, so I have 2GB of data in the virtual memory and another 2GB in the physical memory. Is there a reason why my system should even be using the virtual memory anymore? I would think the computer would run better if it based everything off of RAM instead of virtual memory. Any thoughts on this matter or could you explain why the system is acting this way?"
You must be confused about virtual vs. physical memory. In modern processors, there is no penalty for using virtual memory, all translation from virtual to physical address space is done internal to the processor and you won't notice the difference.
So all the physical memory installed in your PC is used by the processor as one big pool of resources. Processes can think whatever they want and address huge memory spaces, that's all in virtual land. Virtual memory only starts impacting performance when pages are being swapped in and out, because all your processes need more resident memory than you actually have.
Swapping means accessing the disk and freezing the requesting process until its page of memory has arrived from the disk, which takes millions of processor cycles (a lifetime from the processor's point of view). It's not so bad if you swap once, as the processor can work on other processes while waiting for the data to arrive, but if all your programs keep pushing each other out of physical memory, you get thrashing and consider yourself happy if the mouse pointer is still responsive!
So you may want to change the title of your post to: "why use physical memory in modern systems?". I would point you to an article I wrote on that topic in 1990, but somehow I can't find a link to it on the web :-)
fairsoftware.net - software developers share revenue from the apps they build
Memory exists to be used. If memory is not in use, you are wasting it. The reality is that your system will operate with higher performance if unused data is paged out of RAM to disk and the newly freed memory is used for additional disk caching. Vista's memory manager is actually reasonably smart and will only page data out to disk when it really won't be used, or you experience an actual low-memory condition.
Virtual memory and pagefiles still exist so that there will be persistent, recoverable storage of your browsing and search history, illegally downloaded music, and furrie porn should anybody come a-knockin after you hit the power switch.
[/tinfoil hat]
you could create a RAM Disk and set your page file to use that.
Then all your virtual memory is in RAM.
I'll leave it to someone else to explain why that isn't a good idea.
"The stupid neither forgive nor forget; the naive forgive and forget; the wise forgive but do not forget." -Thomas Szasz
We who know what we are doing are free to take the risk of running our computers without a swapfile.
Most people are not in a position where they can be sure that they will never run out of physical memory. Because of that, all operating systems for personal computers set up a swapfile by default: It's better for joe average computer owner to complain about a slow system than for him to lose his document when the system crashes because he filled up the physical memory (and there is no swap file to fall back on).
Stop the brainwash
The other extreme point of view is that modern systems should only have virtual memory and, instead of having an explicit file system, treat mass storage as a level-4 cache. In fact, systems that support mmap(2) do this partially.
The idea here is that modern memory management is actually pretty good, and that it's best to let the OS decide what to keep in RAM and what to swap out, so that issues like prefetching can be handled transparently.
For that matter, why do we even need to explicitely "save" anymore? Why does the fact that Notepad has 2KB of text to save prevent the shutdown of an entire computer? Just save the fecking thing anywhere and get on with it! Modern software is such a disorganized mess.
Absolutely not true. You can even install and run Vista on a computer with 1Gb ram and no page file. And run applications. So it doesn't reserve 1Gb for itself and thus, your myth is busted. Vista's memory manager will use as much memory it can (free memory is a waste, so it will use it rather than watch it empty). But as soon as a process needs memory it will give it back.
It's time to realise that Abble's products are the biggest abomination these days. Just say NO to the dumb iAbble way!!
But, at least in Mac OS X, exited processes consume "inactive" memory - basically being kept in memory until they are launched again. If Vista has a similar implementation, your swapfile may just contain a bunch of pages left over from previously-running applications. Are you experiencing actual system performance problems? Concerning yourself too much with the numbers only can be a bad thing.
I'd assume what he's asking is: in modern systems where the amount of physical RAM is considerably larger than what most people's programs in total use, why does the OS ever swap RAM out to disk?
The answer is basically to free up RAM for disk cache, based on a belief (sometimes backed up by benchmarks) that for typical use patterns, the performance hit of sometimes having to swap RAM back into physical memory is outweighed by the performance gain of a large disk cache.
Of course, OS designers are always revisiting these assumptions---it may be that for some kinds of use patterns using a smaller disk cache and swapping RAM out to disk less leads to better performance, or at least better responsiveness (if that's the goal).
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Virtual memory is now used for little tricks, in addition to providing more memory than is physically available.
One example is ring transitions into kernel mode which start out as exceptions. (Everyone seems to have ignored call gate, the mechanism Intel offered for ring transitions).
Another is memory mapped pointers. It is cool to be able to increment a pointer to file backed ram and not have to care if it is in ram or not.
Maybe the OP is onto something. Imagine writing Windows drivers without having to worry about IRQL and paging.
This should generate some polarized discussion.
There are two camps of thought.
One will insist that, no matter how much memory is currently allocated, it makes more sense to swap out that which isn't needed in order to keep more free physical ram. They will argue until they are blue in the face that the benefits of doing so are good.
Essentially - your OS is clever and it tries pre-emptively swap things out so the memory will be available as needed.
The other camp - and the one I subscribe to - says that as long as you have enough physical ram to do whatever you need to do - any time spent swapping is wasted time.
I run most of my workstations (Windows) without virtual memory. Yes, on occasion, I do hit a "low on virtual memory error" - usually when something is leaky - but I prefer to get the error and have to re-start or kill something rather than have the system spend days getting progressively slower, slowly annoying me more and more, and then giving me the same error.
This is not to say that swap is bad, or that it shouldn't be used - but I prefer the simpler approach.
I can finally put my CS degree to good use, answering the same questions students would ask the TAs in basic OS and systems-level programming courses! ...except that the other comments have already answered the question. So, in true CS fashion, I will be lazy and refrain from duplicating effort ;)
Laziness is a virtue! (And that's on-topic, because a lazy paging algorithm is a good paging algorithm).
I recall back in 2002 or so, a friend of mine maxed out his Windows XP system with 2gb of memory. Windows absolutely refused to turn off paging (swap), forcing him to whatever the minimum size was. The solution? He created a RAMdisk and put the paging file there.
On Linux (and other modern systems, perhaps now including Windows), you can turn off swap. However, the Linux kernel's memory management isn't so great at the situation you hit when you need more memory than you have, but you can't swap. Usually, the memory hog crashes as a result (thankfully, Firefox now has session restore). I might be slightly out of date on this one.
A well-tweaked system still has swap (in nontrivial amounts), but rarely uses it. Trust me, you can afford losing the few gigabytes from your filesystem. Again in Linux, /proc/sys/vm/swappiness can be tweaked to a percentage reflecting how likely the system is to swap memory. Just lower it. (Though note the cons to this presented at the kerneltrap article above.) My workstation currently has an uptime of 14 days, a swappiness of 60, and 42/1427 megs of swap in use as opposed to the 1932/2026 megs of physical memory in use at the moment.
This is summarized for Windows and Linux on Paging at Wikipedia.
Use my userscript to add story images to Slashdot. There's no going back.
thinks "Virtual Memory" is the same thing as paging...
Mac Classic (OS 8 for sure) used the term "Virtual Memory" the same way Windows today uses "Page File" or unix uses "swap", so you can at least understand why some people might be confused by this.
db
I am literally 3000 tokens away from the chaotic crossbow --Stephen
Okay, so we've got most of the "you can run Vista with 4GB?!" jokes out of the way (hopefully). Here's my take on the situation.
...). With your configuration, that's when you'd start to crash.
I have Vista x64 running in a machine with 8GB physical memory, and no page file. I can do this because I'm never running enough memory-hungry processes that I will exceed 8GB allocated memory. So, while the OS may be good at deciding what gets swapped to the hard disk, in my case, there's simply no need, as everything I'm running can be contained entirely within physical memory (and for the curious, I've been running like this for a year and a half, haven't run out of memory yet).
However, if you don't have enough physical memory to store all the processes you might be running at once, then at some point the OS will need to swap to the hard drive, or it will simply run out of memory. I'm honestly not sure exactly how Vista handles things when it runs out of memory (never been a problem, never looked into it), but it wouldn't be good (probably BSoD, crash crash crash). I can tell you from personal experience that I regularly exceed 4GB memory usage (transcoding a DVD while playing a BluRay movie while
Long story short, with just 4GB, I would leave the swap file as is. Really, you should only disable the swap file if you know based on careful observation that your memory usage never exceeds the size of your installed physical memory. If you're comfortable with the risks involved, and you know your system and usage habits well, then go for it. Otherwise, leave it be.
I think he is referring to the userspace/kernelspace split in Windows NT. On 32bit Windows XP, by default, the userspace (ring3) will have at most 2 GB of the physical RAM, and the kernel space would get the rest (some of it paged and some of it not). On systems with more than 3G of RAM (a lot by 2002 standards), it was kinda pointless to reserve that much for the kernel space, so they added a boot.ini flag that changed the split to _AT_MOST_ 3GBytes for the userspace and the rest for kernel space.
In Vista the split for 3G/1G of RAM is default. Actually on a system with 4G of RAM running in 32bit mode, you can't use all of them even if you try (in Windows XP), because right under the 4G limit you would have the PCI memory address mappings, that can be as large as 512M for a common video card with half a gig of RAM. Add to that the RAID controllers and the other hardware, and you have about 800megs of RAM unused because they can't be addressed, as their address-space is used by the installed devices.
I think that http://support.microsoft.com/kb/823440/ and http://support.microsoft.com/kb/171793/ should describe what I'm talking about pretty clearly.
UNIX was not designed to stop you from doing stupid things, because that would also stop you from doing clever ones.
What is left over is the physical memory needed by the system. It seems like the OS preferred a fixed amount of memory, so it would just set up fixed space on the hard disk. So, even if all you have a 1 MB of available memory, the system would set up say 10MB, and that is what would be used. The pages that are being used will be stored in the physical ram, while everything would be stored on the HD.
If page management is working correctly, this should be transparent to the user. The management software or hardware will predict what pages were needed, and transfer those page to ram. One issue we I had was available memory was not hard disk plus physical available ram, but was limited by the available hard disk space.
So, it seems to me that virtual paged memory is still useful because with multiple applications loaded, memory can be a mess, and big fast hard drives it should not be an issue. I don't how Vista works, but it seems that *nix works very hard to insure that the pages that are needed are loaded to physical memory, and page faults do not occur. In this case, where virtual memory equals available physical memory, it would seem that since only physical memory is being used, there would be no performance hit from virtual memory. it is only there in case an application is run that need more memory. It is nice that we do not get those pesky memory errors we got in the very old days.
"She's a scientist and a lesbian. She's not going to let it slide." Orphan Black
Linux kernel maintainer Andrew Morton sets his swappiness to 100 (page as much physical memory as you can, the opposite of this Ask-Slashdot's desires), which he justified in an interview (see above link) by saying:
My point is that decreasing the tendency of the kernel to swap stuff out is wrong. You really don't want hundreds of megabytes of BloatyApp's untouched memory floating about in the machine. Get it out on the disk, use the memory for something useful.
Of course, there's another view, also presented at the above kerneltrap article: If you swap everything, you'll have a very long wait when returning to something you haven't touched in a while.
If you have limited resources, milk the resources you have plenty of; workstations should have high swappiness while laptops, who suffer in disk speed, disk capacity, and power, are probably better suited with lower swappiness. Don't go crazy, though ... swappiness = 0 is the same as running swapoff -a and will crash your programs when they need more memory than is available (as the kernel isn't written for a system without swap).
Use my userscript to add story images to Slashdot. There's no going back.
One word: Multics. Way too far ahead of its time. Those who forget history will have to try to re-invent it. Badly.
It should be. However, people will keep developing bloated programs that require far more RAM then is required to perform their task.
Spelling and Grammar errors have been added to this post for your enjoyment
That page mostly talks about what virtual memory is and doesn't directly list why it is an improvement.
Some folks have already mentioned the fact that it eliminates memory fragmentation, and that it allows mapping of files and hardware into memory without dedicating (wasting) part of the address space to those uses.
Another reason is that you can have 2^64 bytes of total system memory, even if the individual applications are 32-bit, and can only address 2^32 bytes of memory. Since the 32-bit applications are presented a virtual address space, it doesn't matter if their pages are located above the 32-bit boundary.
It means that per-process memory protection is enforced by the CPU paging table. Without virtual memory you would have to reimplement something like it just for memory protection.
It means that the linker/loader don't have to patch the executable with modified address locations when it is loaded into memory.
The above two reasons have the corollary that libraries can be shared in memory much more easily.
And that's just off the top of my head. Virtual memory is a very, very useful thing.
You must be confused about virtual vs. physical memory. In modern processors, there is no penalty for using virtual memory, all translation from virtual to physical address space is done internal to the processor and you won't notice the difference.
Huh? That's totally wrong. If it were true, you wouldn't need any RAM.
It's true that address translation is hard-wired in modern processors. But that just means that figuring out where the data is is as fast as for data that's already in RAM. Actually reading or writing it is only as fast as the media it's stored on. So if you have a lot of big applications running, and there isn't enough RAM for them all to be in physical memory at once, your system "thrashes", as data migrates back and forth between the two media. That's why adding RAM is very often the best way to speed up a slow system, especially if you're running Microsoft's latest bloatware. Defragging the swap disk can also be helpful.
To answer the original question: actually, you often don't need any virtual memory. But sometimes you do. Disk space is cheap, so it makes sense to allocate a decent amount of virtual memory and just not worry about whether it's absolute necessary.
Windows makes me CRAZY about this. the OS is internally configured to use an LRU algorithm to aggressively page.
("Technical bastards" who question my use of paging and swap interchangeably in this post can send their flames to /dev/null \Device\Null or NUL depending on OS)
What I found when disabling paging on an XP pro system with 2GB RAM is that the system performance is explosively faster without the disk IO.
Even an *idle* XP pro system swaps - explaining the time it takes for the system to be responsive to your request to maximize a window you have not used in a while.
I was thrilled to have a rocket-fast system again - until I tried to hibernate my laptop. Note that the hibernation file is unrelated to the swap/paging space.
The machine consistently would blue screen when trying to hibernate if swap/paging was disabled. Enabling swap enabled the hibernation function again. Since reboots take *FOREVER* to reload all the crap that XP needs on an enterprise-connected system - systems mangement, anti-virus agent, software distribution tool, and the required ram-defragger which allows XP to "stand by" when you've got more than 1GB of RAM, plus IM, etc
I reboot as infrequently as possible and consider "stand by" and "hibernate" required functions. As a result, I live with XP and paging enabled, and tolerate the blasted system "unpaging" apps that have been idle a short time.
Poo!
But Herr Heisenberg, how does the electron know when I'm looking?
If it's there, and you can see it . . . it's real.
If you can see it, but it's not there . . . it's virtual.
If you can't see it, and it's not there . . . it's gone.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
The Kessel Run is obviously a surviving salesman problem.
The traveling salesman is selling zombie survival kits at the onset of the zombie apocalypse. He must sell $X worth of kits to afford his choppa ticket, and return to the evac zone. The evac choppa is waiting for him (or does continuous runs), so time is not an issue, and he can make long-winded sales pitches in safe houses.
Distance traveled is an issue, because the horde is everywhere, and the best strategy is to minimize exposure and avoid detection.
Quickness (acceleration, agility) is an issue because it helps you avoid detection, and when detected, you need to escape or hide quickly.
Speed (top speed of your van) is an issue because you often need to make a beeline to the nearest safe house, or to the evac zone once you have met your quota.
A surviving salesman is rated on his total distance traveled. A lower distance is indicative of a better salesman, and a better vehicle. Being able to zoom through the most dangerous areas will shorten your trip (path length) due to the increased demand and reduced supply of zombie survival kits in said areas.
For the Millennium Falcon, the above applies with a few differences. Han Solo and Chewbaca are hiding from the Empire, not the zombie horde. Instead of selling survival kits, they're smuggling contraband. Instead of running to safe houses, they're running off to Mos Eisley or other fringe/pirate friendly planets the Empire doesn't have (complete) control over. The money gained isn't for a choppa ticket, but for the general livelihood of Han and Chewbaca.
I note a lot of people are insisting that "virtual memory" refers to the virtual address space given to a execution context, and what the author really means is "paging".
The funny thing is that these are traditionally poorly defined/understood terms which are gaining a hard consensus for the meanings due to some recent OS books, and poor comp-sci education which insists on a particular definition. Everyone is faulting M$ for using the term incorrectly, even though the original mac OS and other OS's used the term in the same way. Wikipedia defines it one way and then goes on to give historical systems which don't really adhere to the definition. For example the B5000 (considered the first commercial machine with virtual memory) didn't even have "contiguous" working memory as required by the wikipedia definition. It had what would be more specifically called multiple variable sized segments which could be individually swapped. Again, the mac OS evolved from a single process model to muliprocess, in the same address space (look up mac switcher) and implemented "virtual memory" using a system without a MMU by swaping the allocated pieces of memory to disk if they weren't currently locked (in use) and reallocating the memory. Aka they had "virtual memory" in single fragmented address space.
The other example is people use "paging" to describe the act of swaping portions of the memory to disk, misunderstanding that paging is more about splitting an address space or segment up into fixed pieces for address translation to physical, and that disk swapping of pages isn't required for paging. Aka, your system is still "paging" if you disable swapping.
Even the term swapping is unclear because the need to differentiate between swaping pages, and swapping whole processes (or even segments) resulted in people avoided the term swapping to describe systems which were swapping pages instead of segments/regions/processes. These systems were generally called "demand paged" or something similar to indicate that they didn't need to swap a complete process or dataset (see DOSSHELL).
So, give the guy a break, in may ways he is just as correct, if not more so.
Perhaps set up a small ramdisk and pf to that?
"If you have nothing to hide, you have nothing to fear." - Every fascist, ever
Ummm... no. There are a statistically significant number of humans who aren't notably good at anything. I have unhappily encountered too many of them, both in and out of tech work. This is akin to actually believing that "all men are created equal" merely because it would be really really neat and make you feel all warm-and-fuzzy inside if it were true.
Even if your pollyanna perspective was true, being competent at some task doesn't directly equate with an absence of dumbassery. There are numerous species of "dumb" creatures that can be trained to memorize some task and then mimic (repeat) it perfectly ad nauseum... including H. sapiens. An ability to memorize and mimic doesn't equate directly with intelligence. It's a precursor, a prerequisite, perhaps, but not the Real McCoy.
A shocking number of humans, including many regarded as "average" by testing standards, never actually reach a state of true intelligence. Too many of them are profoundly ignorant and quite determined to remain that way.
If you have caches of a size smaller than your real RAM, the order in which you try to access memory really CAN make a difference because cache is many times faster than regular RAM and will try and do things like speculative readahead. If what you are working with is already in the cache by the time you request it then you won't stall for as long.
If you are forever causing the cache to become flushed and forcing the cache to be refilled with a different contents (perhaps because you are causing a large number of random memory access and the cache's readahead is getting your future access wrong so it has to be turned off) then performance will by comparison be slower than a sequential memory access workload.
The above is of course a gross simplification (and doesn't apply if what you are reading fits entirely within cache and is already there). If you have the technical chops you can read more about how order of access can have an impact on speed in Ulrich Drepper's what every programmer should know about memory on LWN.
HKLM\System\CurrentControlSet\Control\Session Manager\Memory Management
DisablePagingExecutive
0 = Drivers and system code can be paged to disk as needed.
1 = Drivers and system code must remain in physical memory.
Ummm... no. There are a statistically significant number of humans who aren't notably good at anything.
to be more precise a lot of people are not good at anything that we know and value, they could be good at stupid things (soccer strategy) and never know it, never make others know and certainly never apply it to something good.
but teoretically anyone could have some bright (but unuseful/unused) capacity.
There are numerous species of "dumb" creatures that can be trained to memorize some task and then mimic (repeat) it perfectly ad nauseum... including H. sapiens. An ability to memorize and mimic doesn't equate directly with intelligence. It's a precursor, a prerequisite, perhaps, but not the Real McCoy.
you forget a prerequisite of intelligence evolution, THE NEED FOR IT.
if they could live their lives in a dumb way many lazy person never try to reach an higer lever of intelligence, never ever feel their lack of it!
A shocking number of humans, including many regarded as "average" by testing standards, never actually reach a state of true intelligence. Too many of them are profoundly ignorant and quite determined to remain that way.
the deepnes of their ignorance is an huge problem because they never perceive their ignorance, and many of the few that perceive it refuse to believe it and assault the condition that make them feel stupid.
that is the sad problem.
It's a FESTIVUS miracle.
You insensitive clod.
Where's my aluminum pole?