Landy Wang - Windows Memory Manager

The Discussion

• 

  I actually asked about exception surfacing, which is distinct from the notion of handling. As poorly stated as the question was, I was really just trying to find out what the MM does when something goes "boo!" way down there as it relates to memory. I guess it makes sense that such a low level system wouldn't package up a Problem and send it up, but certainly, the MM is an aware system, as it relates to memory. Hey, I'm learning too!
  
  C
  

• 

  YAY!  You got Landy!
  
  Landy is one of the two or three Microsoft developers I've interacted with who NEVER gets things wrong.
  
  Which is a remarkable statement if you think about it.  Most people I work with are falible. 
  
  IME, Landy isn't.
  
  

• 

  Very educational!
  
  Good work, thank you.
  

• 

  Interesting and something tells me that Windows would be a lot more stable if everybody moved to managed code.

• 

  koorb wrote:

  Interesting and something tells me that Windows would be a lot more stable if everybody moved to managed code.

  
  
  Three things wrong with that
  
  1. What does it have to do with this topic?
  2. Is Windows unstable?
  3. If your cheap hardware is making Windows unstable how would managed code help that?

• 

  i used to do something like this,
  

       HLOCAL hMem = HeapAlloc(GetProcessHeap(),HEAP_ZERO_MEMORY, HUGE_SIZE);

  ...

       HeapFree(GetProcessHeap(), 0, hMem);
  
  this way is faster than managed code MUCH, and never be wrong.
  

• 

  koorb wrote:

  Interesting and something tells me that Windows would be a lot more stable if everybody moved to managed code.

  
  
  And where would the managed code get it's memory?
  
  Also, you can still leak memory in 100% managed code.
  

• 

  Manip wrote:

  
  
  Three things wrong with that
  
  1. What does it have to do with this topic?
  2. Is Windows unstable?
  3. If your cheap hardware is making Windows unstable how would managed code help that?

  
  
  Like the who video was about memory management. And managed code has its own garbage collector.

• 

  All this talk of food is making my hugry...

• 

  was there longhorn running on the pc in the video?
  

• 

  GRiNSER wrote:

  was there longhorn running on the pc in the video?
  

  
  
  That's a good question. What OS was it, I wonder... 

• 

  Beer28 wrote:

  Orbit86 wrote:could be Longhorn because It was a login screen that I didnt recognize...its linux

  
  
  it's not linux, there's no halt, reboot, ect... bar at the bottom.
  It looks like a lockout screen more than a login screen.
  

  
  
  looks a lot like that
  
  http://www.winsupersite.com/images/review/lh5060_logon.jpg
  

• 

  I'm a little bid sad that they did not go a little bit deeper. Most of the talk was already known from books or the university...

• 

  We'll probably try and go a tad deeper with Landy sometime in the future, addressing the Longhorn Memory Manager, specifically. I realize this wasn't as deep as the other Going Deeps, but the MM is pretty complex and requires more than 30 minutes...
  
  C 

• 

  supersonic wrote:

  
  looks a lot like that
  
  http://www.winsupersite.com/images/review/lh5060_logon.jpg
  

  
  yes it looks exactly like this. [although the picture is offline now, but i saw it when it was there] - i was asking my question because i remembered this picture
  

• 

  Beer28 wrote:

  Are there any major differences between the way windows handles memory pages and the way unix or linux handles pages?
  

  
  
  Yes, lots. I guess, since this is a Going Deep video, that it is time for the obligatory Windows Internals link.
  
  

  Beer28 wrote:

  
  Is there any major differences between the page file and swap?
  Why not have a privilaged swap partition instead of a file?
  

  
  
  NTFS stores everything in files, even the MFT. That way it can move sections of them around if there are corrupt sectors. It also means they can be dynamically resized as and when is necessary.
  
  Again the Window Internals book goes into a lot of detail about this.

• 

  Orbit86 wrote:

  beer, looks like you want to see Windows code...
  

  
  
  We don't need to see any code. There is much scalability between talking about a pagefile that is used to extend memory and the real code.
  
  As beer said he should have talked a little more about the algorithms and how they work.
  
  If you attend a lecture on OS at the university they tell you more then was mentioned in the video. And they simply talk about concepts and not about the actual code.

• 

  Beer28 wrote:

  Is there any major differences between the page file and swap?

  
  
  There is a difference between paging and swapping.  You can't use the two terms interchangeably.
  
  
  

  Beer28 wrote:

  Why not have a privilaged swap partition instead of a file?

  
  
  Pagefile accesses are largely limited by seek times.  This is precisely why Windows tries to put the pagefile in the middle of your partition.
  
  Remember, pagefile access is happening in parallel with other file activity.  The futher the disk heads have to move, the slower your system will run.
  
  If you put your pagefile on a dedicated partition (or use a swap partition like Linux does), then the disk heads have to go out of their way on each pagefile read/write.
  
  The optimal place for a pagefile is in the middle of the most heavily used parition of the hard drive.

• 

  Beer28 wrote:

  Why is swap space on windows used when you have enough physical ram to handle the processes? As opposed to it not being used until it's necessary on linux?

  
  
  There is no such thing as "swap space" in Windows (NT anyway).  Swapping and paging are not the same.  Don't confuse them.
  
  Windows "backs" pages of memory in the pagefile.  I should note that it doesn't back every memory page in the pagefile.  Things like EXE and DLL code are already backed in their on-disk files.
  
  Notice that I used the word "backs".  The pagefile is (loosely) mirroring the contents of RAM.  This means that when an app does need to get paged to disk, it's already in the pagefile.  The RAM can just be marked as free.
  
  If this wasn't done, when an app needed to be paged out, Windows would have to write the modified pages to disk right then and there.  This would increase the disk thrashing when a system gets maxed out.
  
  Yes Windows is using the pagefile.  Is it actually running programs from the pagefile?  Probably not.  Unfortunately, a lot of people don't understand this.
  
  I feel that Task Manager is to blame for some of this.  The way that it reports pagefile usage is misleading.  If you want to get an idea for how much of your pagefile is really being used, run Perfmon.
  
  FYI, You are incorrect when you say that Linux doesn't use the swap space with lots of physcial memory.  It does something very similar to what I have described above.

• 

  I haven't tested whether there's a real perf benefit for this, but have always thought that it's optimal to have one drive/array for system and apps and second for pagefile and backups/infrequently accessed data. I am talking about what is the optimal 2 hdd setup for both performance and some redundancy - special arrangement like this or some special RAID that gives both perf + redundancy on just 2 drives?
  

• 

  androidi wrote:

  I haven't tested whether there's a real perf benefit for this, but have always thought that it's optimal to have one drive/array for system and apps and second for pagefile and backups/infrequently accessed data.

  
  
  Assuming you have multiple drives, the best place for the pagefile is the middle of the most-used partition on the least-used drive.
  

• 

  well the pic isn't offline , its still there
  
  if u can see it from the direct link , go to that site
  http://www.winsupersite.com/showcase/longhorn_preview_2005.asp
  
  scroll down a bit und u will see the last thunbnailpic on the right sight
  
  
  

• 

  msemack wrote:

  
  Assuming you have multiple drives, the best place for the pagefile is the middle of the most-used partition on the least-used drive.
  

  
  
  The best place is the only partition on a dedicated drive on a dedicated controller. Anything else is a compromise of some sort.
  
  And even that's assuming all the drives are similar speeds. Which may not be the case.

• 

  AndyC wrote:

  The best place is the only partition on a dedicated drive on a dedicated controller. Anything else is a compromise of some sort.

  
  
  I agree.  And that is fully compatible with what I said.  It's the most-used partition (the only one) of the least-used drive. 
  
  

  AndyC wrote:

  And even that's assuming all the drives are similar speeds. Which may not be the case.

  
  
  This is true.  However, in a system with a sufficient amount of RAM, the pagefile will be hit so infrequently that this all becomes academic.
  

• 

  Beer28 wrote:

  
  Swap is always at 0 on my system monitor on FC4 with about a gig of ram. top reports it at 0 as well.
  
  The only time I remember seeing it at anything other than 0 was when I only had 256 MB of ram a long time ago. And on my server where I only have 512MB.
  
  

  
  
  Reporting a 0 doesn't necessarily mean your swap space is completely unused.
  
  Edit: http://sourcefrog.net/weblog/software/linux-kernel/swap.html

• 

  I have a basic doubt. Whether memory to memory transfer using system DMA controller is possible in windows?
  If not why?
  Please reply back me on sonypappan@yahoo.com or sonypappan@hotmail.com
  Regards,
  Sony

• 

  Beer28 wrote:

  Does the memory manager work with the kernel scheduler?
  
  If a process releases it's timeslice as soon as it's started because it's a service or never accesses a good portion of memory or is in a sleep wait state, is it's heap more likely to be paged out?
  
  What about non-paged driver memory? How does the memory manager handle that? ISR's can't use paged memory, are they still handled by the memory manager on windows TM?
  

  
  
  Heap is always most likely to paged out. That's why we have working-sets to keep processes happy.
  
  ISRs cannot use paged memory as a general rule, and a good one if you care about performance. No fun waiting for the file system to respond in at least 3 orders of magnitude delays. With that said, non-paged memory is simple: Don't use it if you don't absolutely need it. These are non-paged and thusly subtracted from your total free pool. When you unload your driver, all the allocated memory goes back to the system.
  
  The memory manager only cares about allotting free pages and reclaiming pages that are committed, or pages from working-sets. People here are confusing paging for swapping and vice versa. Swapping is swapping a good portion of memory to disk to free memory. Paging is swapping out a "page" of memory, in i386 world that's typically 4KB.
  
  By the kernel scheduler, you mean thread scheduler? These are two different concepts. The VMM probably implements LRU or something of that sort or use a clock algorithm. Who knows? VMM typically has nothing to do with kernel scheduling. Should we entertain this idea,  the prediction of when memory resources should be freed for other proesses is probably left to some kind of cache policy manager. That's a big research task! For now the working-set is our best bet for temporal and local cache prediction without requiring us to think too hard about when to actually free memory with a scheduler. We can easily reclaim memory from the LRU working-set pages from any process, or randomly if we want pseudo-LRU characteristics.
  

• 

  androidi wrote:

  I haven't tested whether there's a real perf benefit for this, but have always thought that it's optimal to have one drive/array for system and apps and second for pagefile and backups/infrequently accessed data. I am talking about what is the optimal 2 hdd setup for both performance and some redundancy - special arrangement like this or some special RAID that gives both perf + redundancy on just 2 drives?
  

  
  
  Haha you're fanny! Just add more memory and avoid page files altogether. There's a product like SuperCache that does Disk I/O level caching. All of course at the expense of reserving more memory for these operations. Personally I gave up page files after disk thrashing really put a burden on performance, especially when we're talking about compiling and linking 800MB worth of source codes every 15 minutes. The only reason you get "more" performance boost by putting the paging file on the second drive is that whenever you access your system the first disk is busy and the second drive is free to work. If you really think about this, to really get the benefits, at least for IDE "technology", you needed to put each device on its own separate channel. Basically one disk per ribbon. But then again, reality hits you again. We're talking about many orders of magnitudes of slowness because the disk is just so slow. You should look into borrowing memory from other systems over the network, if you're kinda cheap, since accessing them is still much faster than your local disks.
  

• 

  Hi there:

  I´m a end user from Venezuela in the multimedia field, (using 3dmax, premiere, AfterFX, photoshop and so on), with a basic survival knowledge of windows set up. Mi system is a custom ordered Pentium D on Asus MB with 2 gig of ram, and lots of disk space. As you may know, all the applications I use are memory-cannibal, so I don´t know if I can disable the swap memory without affecting the performance of the programs. It´s there some sort of a program that "cleans" the swap memory after exiting an application without having to restart windows? I ask this because some times after a heavy use of some of this programs, the performance of the computer comes down to a crawl, which it´s noticeable in the update of the widows in the win explorer, for example.
  
  Thanks in advance for any advice on this issue.

• 

  Beer28 wrote:

  
  ...
  I can't help but notice that even though swap is enabled on linux, it's hardly ever used because I have alot of ram. It's not turned off. But linux will not use the disk swap partition space unless it's necessary. I can have 1 gig of ram on windows and have notepad open and it will still use the swap file if I have it enabled. It obviously doesn't need it, why is that?
  
  Why is swap space on windows used when you have enough physical ram to handle the processes? As opposed to it not being used until it's necessary on linux?
  

  
  
  This is actually really crappy behavior on the part of Linux.  In my experience, in real load scenarios, the behavior of swapping out entire processes only when under heavy load tends to lead to trashing.  The system spends most of its time swapping out entire processes only to have to struggle to swap them back in so they can get their share of CPU time, to only then swap them back out again, etc.
  
  Lots of systems, not just Windows, will page out unused pages ahead of time.  In certain cases (like heap pages which tend to be backed by swap), this paging out is pushed to swap.  So it's normal to see swap in use even though there are lots of free pages hanging around (i.e., lots of free RAM) -- on FreeBSD, or similar, this is what happens.
  
  -B
  

• 

  So if the VMM is so smart, when will Internet Explorer NOT leak memory? One of the most aggravating things about IE is that it steadily expands to fill all available memory, degrading the system over time. This is true for IE7 and IE8 and is clearly observable in Task Manager. Exiting all instances of IE doesn't free it all and eventually the only solution is to reboot.

• 

  erm 
  "Posted: 22 hours ago" or a few years ago ?

  what's up with reposting old content ? Did some of the converting process fail (old -> new site) ?

  Or are you just lacking new content to upload ? 

• 

  Great interview and a great personality, thanks for that! Interview would have been better with a bit of upfront preparation in terms of questions.