Too Much Cache?
Cache is used to reduce the performance impact when accessing data that resides on slower storage media. Without it your PC would crawl along and become nearly unusable. If data or code pages for a file reside on the hard disk, it can take the system 10 milliseconds to access the page. If that same page resides in physical RAM, it can take the system 10 nanoseconds to access the page. Access to physical RAM is about 1 million times faster than to a hard drive. It would be great if we could load up all the contents of the hard drive into RAM, but that scenario is cost prohibitive and dangerous. Hard disk space is far less costly and is non-volatile (the data is persistent even when disconnected from a power source).
Since we are limited with how much RAM we can stick in a box, we have to make the most of it. We have to share this crucial physical resource with all running processes, the kernel and the file system cache. You can read more about how this works here:
https://blogs.msdn.com/ntdebugging/archive/2007/10/10/the-memory-shell-game.aspx
The file system cache resides in kernel address space. It is used to buffer access to the much slower hard drive. The file system cache will map and unmap sections of files based on access patterns, application requests and I/O demand. The file system cache operates like a process working set. You can monitor the size of your file system cache's working set using the Memory\System Cache Resident Bytes performance monitor counter. This value will only show you the system cache's current working set. Once a page is removed from the cache's working set it is placed on the standby list. You should consider the standby pages from the cache manager as a part of your file cache. You can also consider these standby pages to be available pages. This is what the pre-Vista Task Manager does. Most of what you see as available pages is probably standby pages for the system cache. Once again, you can read more about this in "The Memory Shell Game" post.
Too Much Cache is a Bad Thing
The memory manager works on a demand based algorithm. Physical pages are given to where the current demand is. If the demand isn't satisfied, the memory manager will start pulling pages from other areas, scrub them and send them to help meet the growing demand. Just like any process, the system file cache can consume physical memory if there is sufficient demand.
Having a lot of cache is generally not a bad thing, but if it is at the expense of other processes it can be detrimental to system performance. There are two different ways this can occur - read and write I/O.
Excessive Cached Write I/O
Applications and services can dump lots of write I/O to files through the system file cache. The system cache's working set will grow as it buffers this write I/O. System threads will start flushing these dirty pages to disk. Typically the disk can't keep up with the I/O speed of an application, so the writes get buffered into the system cache. At a certain point the cache manager will reach a dirty page threshold and start to throttle I/O into the cache manager. It does this to prevent applications from overtaking physical RAM with write I/O. There are however, some isolated scenarios where this throttle doesn't work as well as we would expect. This could be due to bad applications or drivers or not having enough memory. Fortunately, we can tune the amount of dirty pages allowed before the system starts throttling cached write I/O. This is handled by the SystemCacheDirtyPageThreshold registry value as described in Knowledge Base article 920739: https://support.microsoft.com/default.aspx?scid=kb;EN-US;920739
Excessive Cached Read I/O
While the SystemCacheDirtyPageThreshold registry value can tune the number of write/dirty pages in physical memory, it does not affect the number of read pages in the system cache. If an application or driver opens many files and actively reads from them continuously through the cache manager, then the memory manger will move more physical pages to the cache manager. If this demand continues to grow, the cache manager can grow to consume physical memory and other process (with less memory demand) will get paged out to disk. This read I/O demand may be legitimate or may be due to poor application scalability. The memory manager doesn't know if the demand is due to bad behavior or not, so pages are moved simply because there is demand for it. On a 32 bit system, the file system cache working set is essentially limited to 1 GB. This is the maximum size that we blocked off in the kernel for the system cache working set. Since most systems have more than 1 GB of physical RAM today, having the system cache working set consume physical RAM with read I/O is less likely.
This scenario; however, is more prevalent on 64 bit systems. With the increase in pointer length, the kernel's address space is greatly expanded. The system cache's working set limit can and typically does exceed how much memory is installed in the system. It is much easier for applications and drivers to load up the system cache with read I/O. If the demand is sustained, the system cache's working set can grow to consume physical memory. This will push out other process and kernel resources out to the page file and can be very detrimental to system performance.
Fortunately we can also tune the server for this scenario. We have added two APIs to query and set the system file cache size - GetSystemFileCacheSize() and SetSystemFileCacheSize(). We chose to implement this tuning option via API calls to allow setting the cache working set size dynamically. I’ve uploaded the source code and compiled binaries for a sample application that calls these APIs. The source code can be compiled using the Windows DDK, or you can use the included binaries. The 32 bit version is limited to setting the cache working set to a maximum of 4 GB. The 64 bit version does not have this limitation. The sample code and included binaries are completely unsupported. It is just a quick and dirty implementation with little error handling.
Comments
Anonymous
November 27, 2007
I'm getting "is not a valid Win32 application" error. [I’m guessing that you are running SetCache.exe on Windows XP (or earlier). The GetSystemFileCacheSize and SetSystemFileCacheSize API functions require Windows Server 2003 SP1 or later (this includes Windows XP x64 Edition, since it is built from the 2003 SP1 codebase). Since these functions don’t exist on earlier versions of Windows, the SetCache.exe binary was compiled with the subsystem version set to 5.02, which prevents it from running on versions of Windows where the API functions do not exist.]Anonymous
January 13, 2008
Good article. keep it up to keep us up to date.Anonymous
January 14, 2008
The comment has been removedAnonymous
March 26, 2008
how do u find out how much cache is on your pc? (nice article learnt a bit keep i up) [If you want to see how much cache is currently being used, you can do so with Performance Monitor counter /Memory/Cache Bytes. If you want to see the limits, you can use the SetCache executable or !filecache in a postmortem or live debug. !filecache will also show you the current file cache usage along with how much cache each file is using.]Anonymous
May 08, 2008
The comment has been removedAnonymous
May 14, 2008
Fantastic! I have upgraded our backup server to W2k8x64 from w2k3x86 and it kept using all physical RAM, I even put 16GB in it and it used it all! this explains it all! I have managed to reduce the cache to 1GB, but can't set it at more if I try anything over 1GB it sets to 8TB, e.g.: C:>setcache 2048 Current Cache Settings: Minimum File Cache Size: 100 MBytes Maximum File Cache Size: 1024 MBytes Flags: 1 New Cache Settings: Minimum File Cache Size: 100 MBytes Maximum File Cache Size: 8388607 MBytes Flags: 1 Any ideas? otherwise I'll take some of this expensive memory out if it can't be used usefully! [Thanks for the feedback. It turns out there was a bug in the sample code. The bug reveals that I didn't have a modern system with a lot of RAM to properly test this code. I've updated the code and the binaries. Try this new version.]Anonymous
June 18, 2008
The comment has been removedAnonymous
July 05, 2008
The comment has been removedAnonymous
August 29, 2008
The comment has been removedAnonymous
September 07, 2008
Excessive paging on Exchange 2007 servers when working sets are trimmedAnonymous
October 03, 2008
I used setCache.exe it on win2008 64bit enterprise to 2048MB and the "cached" value that shows in task manager (directly below memory guage) just keeps going higher than 2048MB. Mine is now 12,224MB! Does anyone know what is going on? [Task Manager’s value for “Cached” is not what you think it is. In addition to the Cache Manager’s working set, this number includes the number of standby pages in physical RAM. While standby pages are like cached pages, they can be quickly disassociated with the previous working set, scrubbed and handed to a new process. Task Manager is just showing you another way of looking at the data. To see the real working set size of the System File Cache you need to use Performance Monitor. You can read more about this in the Memory Shell Game post.]Anonymous
December 16, 2008
Using the sample code, is it possible to set the minimum cache size? It defaults to 100MB instead of the original 1MB. [By default the sample code is hard coded to set the minimum to 100 MB, but not enforce it. This was done to allow the memory manager to reduce the working set size as needed. You can modify the sample code to set a hard limit by changing the Flags parameter in the call to SetSystemFileCacheSize().]Anonymous
December 16, 2008
This reads and acts like a Russinovich post and tool - easily understandable, educational, small, fast, and useful. Thank you! I have Vista Home Premium x64 with 4GB RAM. 1GB max cache on a 4GB system seems much more reasonable than the default of 8.4 TB! Do I need to reboot in order for the new cache setting to take effect? [Thanks for the feedback. In response to your question a reboot is not required. The setting is dynamically applied to the system file cache’s working set size. It is important to note this setting is not persistent so rebooting the machine will revert the size back to the default. In order to maintain the settings you’ll need to run the tool at least once per boot. One approach is the use of a machine start up script to automate the process after a reboot.]Anonymous
February 06, 2009
Excessive cached read I/O is a growing problem. For over one year we have been working on this problemAnonymous
February 07, 2009
Rilasciato Microsoft Windows Dynamic Cache ServiceAnonymous
March 11, 2009
Tool works fine on 2008 x64 in that it commits the change immediately after running the command. However the setting doesn't stick after a server reboot but defaults back to 8386607MB. Is there a way to make the setting permanent or do we have to resolve to running the tool during the Windows startup sequence? [Good question. SetCache has been replaced by the Microsoft Windows Dynamic Cache Service. You read more about it here. These settings are not persistent and will revert to default values when the system starts. You can either use SetCache in a local system startup script or use the Microsoft Windows Dynamic Cache Service. The new service will auto-start with the system and set the cache limit based on many configurable options.]Anonymous
April 22, 2009
Thank you very much for your post. I am so sory not found it 2 years ago... I am trying to donload binaries with no success. Tried several ISP's. Please, help?Anonymous
June 08, 2009
CecoM, this has now been replaced with Microsoft Windows Dynamic Cache Service - grab it from http://www.microsoft.com/downloads/details.aspx?FamilyID=e24ade0a-5efe-43c8-b9c3-5d0ecb2f39af&displaylang=enAnonymous
October 13, 2009
I just tried to install this tool on Windows 2008 R2 but it failed to start with the notification that the tool was written for an earlier version of Windows. Will there be an update soon or is there a tool provided within R2 that manages the cache size? [ Thanks for the great question! The SetCache tool is not an installable tool. Additionally it has been replaced by the Microsoft Windows Dynamic Cache Service. Either way, these tools help to mitigate the problem of excessive growth of the system file cache on versions of Windows prior to Windows 7 and Windows Server 2008 R2. We have updated the memory manager algorithms in Windows 7 and Windows Server 2008 R2 to address this issue natively in the Operating System. You should not use these tools on the latest version of Windows. ]Anonymous
March 06, 2010
The comment has been removedAnonymous
July 06, 2010
I'm a SQL Server Analysis Services MVP, and I'm very interested in the interaction between the system file cache and SSAS (as it leverages the system file cache heavily). Do you know any experts on that topic from Microsoft that I should ping? Two quick questions for you. I've written some C# code that lets you clear the windows system file cache. The main reason for this is to repeatably retest the performance of an Analysis Services query on a completely cold system file cache (without server reboot). Two questions: 1. Is uses NtSetSystemInformation. Is this doing anything different than SetSystemFileCacheSize? You can see the code here: asstoredprocedures.svn.codeplex.com/.../FileSystemCache.cs 2. From your article, it appears that limiting the system file cache doesn't zero the system file cache memory that's trimmed, but rather moves it to standby. Is there an API (or any other way) of clearing standby memory in the system file cache? That code mentioned above isn't producing repeatable cold system file cache tests, and I believe it's because soft faults from standby are so much faster than hard faults. [SetSystemFileCacheSize() internally uses NtSetSystemInformation(). I would recommend using SetSystemFileCacheSize() over NtSetSystemInformation() because SetSystemFileCacheSize() is a public API. While you can use NtSetSystemInformation(), you run a greater risk of your application breaking if we change the interface. If you want to clear out physical RAM without a server reboot, I would recommend creating an application that will consume most of available memory and then dump the pages onto the free list. First get the current File System Cache’s working set size, then set the limit to something fairly low (but not too low). Next find out how much available memory is on the system, and then allocate that much memory in your process (leave about 64 MBs free). Write at least one byte per page to guarantee that it will be committed to your process’s working set. Finally restore the System File Cache’s work set to where it was and then exit the process.]Anonymous
October 11, 2011
I'm not able to start the DynCache service on my Windows Server 2008 SP2 64-bit. Struggling with these for some time now. Please help!!!
Services
Windows could not start the Dynamic Cache Service service on Local Computer. Error 216: 0xd8
OK
Anonymous
November 19, 2011
On Windows 7 SP1 I had to hardcode into the LimitCache function: if(MaxCacheSize>51210241024) MaxCacheSize=51210241024; and it now works:) (upper limit is 512MBytes, as reported by SysInternals CacheSet, and once it reached this limit, it didn't cross it).Anonymous
March 19, 2012
Hello, very interesting article. I am using Windows Server 2008 R2 and am seeing this runaway file cache issue consuming all of the available physical RAM. My application does a ton of random access reads and writes. What were the changes to the memory manager between Windows 2008 and 2008 R2? I am curious since the runaway cache problem is still there. What is your recommendation for dealing with it on Windows 2008 R2? [The cache manager in Windows Server 2008 R2 handles almost all scenarios more efficiently, and usually avoids the need for dyncache. Unless you have many individual files open, the cache manager should not encounter the scenario described in this article on R2. It is difficult to provide 1:1 support through blog comments, if you need troubleshooting assistance you may want to open a support incident so that our engineers can assist you.]Anonymous
December 20, 2012
Ugh, this is terribad. With the current Steam sale lots of downloading is being done. While Steam is downloading 'Cache WS' in Process Explorer is constantly growing. I saw 3GB Cache WS on the 4GB system and instead of discarding this clearly useless memory, it's starting to swap running programs to disk, aarrgh. It's currently so bad I just run Cacheset.exe 1024 1024 half-hourly. Not that it would be stick to anywhere near 1024KB but at least it clears the cache instantly. [You may benefit from the service described in this article: http://blogs.msdn.com/b/ntdebugging/archive/2009/02/06/microsoft-windows-dynamic-cache-service.aspx.]Anonymous
February 26, 2013
Is SystemCacheDirtyPageThreshold still relevant for Windows Server 2012? Thanks! [That setting is only relevant on Windows Server 2003 SP2, or SP1 with KB920739 installed. For more information refer to http://support.microsoft.com/kb/920739.]Anonymous
March 13, 2014
On one of your posts I see the below response- { Thanks for the great question! The SetCache tool is not an installable tool. Additionally it has been replaced by the Microsoft Windows Dynamic Cache Service. Either way, these tools help to mitigate the problem of excessive growth of the system file cache on versions of Windows prior to Windows 7 and Windows Server 2008 R2. We have updated the memory manager algorithms in Windows 7 and Windows Server 2008 R2 to address this issue natively in the Operating System. You should not use these tools on the latest version of Windows. ] However I am experiencing the MetaFile utilizing over 90% of my RAM on Windows Server 2008 R2 w/ SP1 that has 16GB of RAM, should I install the Windows Dynamic Cache Services? - www.microsoft.com/.../details.aspx I also question this as the ReadMe file included the the Dynamic Cache Service zip states the below - This service will only run on Windows Server 2008 R2 or earlier versions of Windows. Do not attempt to run this service on a version of Windows after Windows Server 2008 R2 as it will most likely cause performance problems. [Please refer to this article for information on dyncache: http://blogs.msdn.com/b/ntdebugging/archive/2009/02/06/microsoft-windows-dynamic-cache-service.aspx]Anonymous
May 01, 2014
I'm trying to run this service on Windows Server 2008 R2 (64 bit) but have been unable to make it work. Should I use the the Dyncache.exe under the I386 folder or the one under AMD64? If I use the 64 bit one, should that go under the C:WindowsSysWOW64 or should it be under the System32 folder? [The answer to your question depends on which 64-bit version of Windows you are using. If you are using x64, then use the version in the amd64 folder and put it in the system32 folder. Note that AMD64 refers to the 64-bit extended x86 architecture, and is not specific to hardware manufactured by AMD. If you are using the Itanium version of Windows, use the version in the ia64 folder and put it in system32.]Anonymous
September 11, 2014
The link to the source code seems to be dead. [This tool is obsolete and was replaced by dyncache. The source for dyncache is included in the download package. http://blogs.msdn.com/b/ntdebugging/archive/2009/02/06/microsoft-windows-dynamic-cache-service.aspx]Anonymous
October 03, 2014
What it means: The system clock varies too much (v=662.91) [This text does not appear in the article, unfortunately we do not know what it means either.]Anonymous
December 02, 2014
RegValue: MaxSystemCacheMBytes Type: REG_DWORD Values: 0 = Limit to 90% of Physical RAM (default) 1-99 = Limit the maximum size of the System File Cache to this percentage of Physical RAM > 200 = Limit the maximum size of the System File Cache to x Mbytes How this works: This setting is the absolute maximum that the System File Cache’s working set could be set to. The default is 0, limiting it to 90% of physical RAM with an upper limit of total Physical RAM minus 300 Mbytes. The lower limit for absolute values is 200 Mbytes and it must be at least 100 Mbytes greater than the MinSystemCacheMBytes value, which defaults to 100 Mbytes. [It is good to see someone is reading the readme file.]