Microsoft MVP성태의 닷넷 이야기
.NET Framework: 492. .NET CLR Memory 성능 카운터의 의미 [링크 복사], [링크+제목 복사],
조회: 22239
글쓴 사람
정성태 (techsharer at outlook.com)
홈페이지
첨부 파일

(시리즈 글이 12개 있습니다.)
Windows: 103. 작업 관리자에서의 "Commit size"가 가리키는 메모리의 의미
; https://www.sysnet.pe.kr/2/0/1850

.NET Framework: 492. .NET CLR Memory 성능 카운터의 의미
; https://www.sysnet.pe.kr/2/0/1852

.NET Framework: 626. Commit 메모리가 낮은 상황에서도 메모리 부족(Out-of-memory) 예외 발생
; https://www.sysnet.pe.kr/2/0/11110

VC++: 107. VirtualAlloc, HeapAlloc, GlobalAlloc, LocalAlloc, malloc, new의 차이점
; https://www.sysnet.pe.kr/2/0/11152

Windows: 136. Memory-mapped File은 Private Bytes 크기에 포함될까요?
; https://www.sysnet.pe.kr/2/0/11159

.NET Framework: 845. C# - 윈도우 작업 관리자와 리소스 모니터의 메모리 값을 구하는 방법
; https://www.sysnet.pe.kr/2/0/11950

Windows: 211. Windows - (commit이 아닌) reserved 메모리 사용량 확인 방법
; https://www.sysnet.pe.kr/2/0/13133

.NET Framework: 2073. C# - VMMap처럼 스택 메모리의 reserve/guard/commit 상태 출력
; https://www.sysnet.pe.kr/2/0/13174

.NET Framework: 2074. C# - 스택 메모리에 대한 여유 공간 확인하는 방법
; https://www.sysnet.pe.kr/2/0/13180

Linux: 57. C# - 리눅스 프로세스 메모리 정보
; https://www.sysnet.pe.kr/2/0/13221

닷넷: 2322. C# - 프로세스 메모리 중 Private Working Set 크기를 구하는 방법(성능 카운터, WMI)
; https://www.sysnet.pe.kr/2/0/13889

닷넷: 2323. C# - 프로세스 메모리 중 Private Working Set 크기를 구하는 방법(Win32 API)
; https://www.sysnet.pe.kr/2/0/13890




.NET CLR Memory 성능 카운터의 의미

다음의 글에 나오는 성능 카운터를 한번 볼까요? ^^

Memory Performance Counters
; https://learn.microsoft.com/ko-kr/previous-versions/x2tyfybc(v=vs.110)

그중에서 먼저 아래의 성능카운터를 살펴보겠습니다.

# Total committed Bytes: Displays the amount of virtual memory, in bytes, currently committed by the garbage collector. Committed memory is the physical memory for which space has been reserved in the disk paging file.

# Total reserved Bytes: Displays the amount of virtual memory, in bytes, currently reserved by the garbage collector. Reserved memory is the virtual memory space reserved for the application when no disk or main memory pages have been used.


지난 글을 보셨다면,

작업 관리자에서의 "Commit size"가 가리키는 메모리의 의미
; https://www.sysnet.pe.kr/2/0/1850

이제 "# Total committed Bytes"와 "# Total reserved Bytes"의 차이점을 아실 수 있을 것입니다. 전자는 실제로 메모리/디스크에 할당이 된 크기이고, 후자는 주소 공간 예약까지를 포함한 크기입니다. 따라서, 이름에서 알 수 있듯이 ".NET CLR Memory" 범주의 "# Total committed Bytes", "# Total reserved Bytes" 성능 카운터가 가리키는 수치는 '가상 메모리' 중에서도 "CLR"에 의해 할당된 크기만을 나타냅니다. 즉, GC Heap으로 인한 가상 메모리 예약/할당 크기입니다.

코드에서 다음과 같이 수치를 구하면,

Process process = System.Diagnostics.Process.GetCurrentProcess();
Console.WriteLine((process.VirtualMemorySize64 / 1024).ToString("#,#"));

VirtualMemorySize64 숫자는 언제나 ".NET CLR Memory" 범주의 "# Total reserved Bytes" 수치보다 큽니다. (참고로, 성능 카운터에 VirtualMemorySize64와 동일한 크기를 나타내는 수치로 Process 범주의 "Virtual Bytes"가 해당합니다.)




그다음 GC Heap 메모리 쪽의 수치를 볼까요?

# Bytes in all Heaps: Gen 1 Heap Size, Gen 2 Heap Size, and Large Object Heap Size counters This counter indicates the current memory allocated in bytes on the garbage collection heaps.

Gen 0 heap size: Displays the maximum bytes that can be allocated in generation 0; it does not indicate the current number of bytes allocated in generation 0.

Gen 1 heap size: Displays the current number of bytes in generation 1; this counter does not display the maximum size of generation 1. Objects are not directly allocated in this generation; they are promoted from previous generation 0 garbage collections. This counter is updated at the end of a garbage collection, not at each allocation.

Gen 2 heap size: Displays the current number of bytes in generation 2. Objects are not directly allocated in this generation; they are promoted from generation 1 during previous generation 1 garbage collections. This counter is updated at the end of a garbage collection, not at each allocation.

Large Object Heap size: Displays the current size, in bytes, of the Large Object Heap. Objects that are greater than approximately 85,000 bytes are treated as large objects by the garbage collector and are directly allocated in a special heap; they are not promoted through the generations. This counter is updated at the end of a garbage collection, not at each allocation.


순간적인 값을 재었을 때,

# Bytes in all Heaps:           61,200,000

-------------------------------------------
Gen 0 heap size:		 3,145,728
-------------------------------------------
Gen 1 heap size:		 2,328,000
Gen 2 heap size:		24,272,000
Large Object Heap size:         34,600,000
          (Gen1 + Gen2 + LOH == 61,200,000)

문서의 내용에 나온 것처럼, "# Bytes in all Heaps" 수치는 정확히 (Gen0을 제외한) "Gen1 + Gen2 + LOH Heap"의 크기와 일치합니다. 아마도 Gen 0 힙에서는 빈번하게 메모리 할당이 이뤄질 것이기 때문에 그 순간의 수치가 크게 의미가 없다고 판단한 것이 아닌가 생각됩니다. 그 외의 수치(Gen1, Gen2, LOH)는 GC가 발생하지 않으면 고정이기 때문에 의미가 있고.

그런데 "Gen 0 heap size"의 숫자가 왠지 낯설지 않습니다. 왜냐하면 3 * 1024 * 1024로 얻을 수 있는 수치로 정확히 3MB에 해당하기 때문입니다. 즉, Gen 0 힙 크기는 항상 3MB를 유지하면서 간다고 짐작할 수 있습니다.

(참고로, 여기서 밝히는 모든 수치는 상황에 따라 달라질 수 있습니다. 가령, 메모리 할당 상황에 따라 Gen 0 힙 크기가 3MB가 아닐 수도 있고, 닷넷 버전에 따라 달라질 수 있습니다.)




이와 같은 성능 카운터에 해당하는 수치를 코드로 구하는 것이 가능할까요? 물론입니다. ^^ 이를 위해서는 약간 복잡한 Inteop 코드를 포함해야 하는데요. 다음과 같이 고정적인 인터페이스 및 구조체 정의가 필요합니다.

[Flags]
public enum COR_GC_STAT_TYPES
{
    COR_GC_COUNTS = 0x1,
    COR_GC_MEMORYUSAGE = 0x2
}

[StructLayout(LayoutKind.Sequential, Pack = 4)]
public struct _COR_GC_STATS
{
    public uint Flags;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint ExplicitGCCount;
    [MarshalAs(UnmanagedType.ByValArray, SizeConst = 3)]
    public uint[] GenCollectionsTaken;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint CommittedKBytes;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint ReservedKBytes;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint Gen0HeapSizeKBytes;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint Gen1HeapSizeKBytes;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint Gen2HeapSizeKBytes;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint LargeObjectHeapSizeKBytes;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint KBytesPromotedFromGen0;
    [ComAliasName("mscoree.ULONG_PTR")]
    public uint KBytesPromotedFromGen1;
}

[StructLayout(LayoutKind.Sequential, Pack = 8)]
public struct _COR_GC_THREAD_STATS
{
    public ulong PerThreadAllocation;
    public uint Flags;
}

[ComImport, InterfaceType((short)1), Guid("FAC34F6E-0DCD-47B5-8021-531BC5ECCA63")]
public interface IGCHost
{
    void SetGCStartupLimits([In] uint SegmentSize, [In] uint MaxGen0Size);
    void Collect([In] int Generation);
    void GetStats([In, Out] ref _COR_GC_STATS pStats);
    void GetThreadStats([In] ref uint pFiberCookie, [In, Out] ref _COR_GC_THREAD_STATS pStats);
    void SetVirtualMemLimit([In, ComAliasName("mscoree.ULONG_PTR")] uint sztMaxVirtualMemMB);
}

[ComImport, InterfaceType((short)1), Guid("F31D1788-C397-4725-87A5-6AF3472C2791")]
public interface IGCThreadControl
{
    void ThreadIsBlockingForSuspension();
    void SuspensionStarting();
    void SuspensionEnding(uint Generation);
}

[ComImport, Guid("5513D564-8374-4CB9-AED9-0083F4160A1D"), InterfaceType((short)1)]
public interface IGCHostControl
{
    void RequestVirtualMemLimit([In] uint sztMaxVirtualMemMB, [In, Out] ref uint psztNewMaxVirtualMemMB);
}

[ComImport, InterfaceType((short)1), Guid("23D86786-0BB5-4774-8FB5-E3522ADD6246")]
public interface IDebuggerThreadControl
{
    void ThreadIsBlockingForDebugger();
    void ReleaseAllRuntimeThreads();
    void StartBlockingForDebugger(uint dwUnused);
}

[ComImport, Guid("5C2B07A5-1E98-11D3-872F-00C04F79ED0D"), InterfaceType((short)1)]
public interface ICorConfiguration
{
    void SetGCThreadControl([In, MarshalAs(UnmanagedType.Interface)] IGCThreadControl pGCThreadControl);
    void SetGCHostControl([In, MarshalAs(UnmanagedType.Interface)] IGCHostControl pGCHostControl);
    void SetDebuggerThreadControl([In, MarshalAs(UnmanagedType.Interface)] IDebuggerThreadControl pDebuggerThreadControl);
    void AddDebuggerSpecialThread([In] uint dwSpecialThreadId);
}

[ComImport, ComConversionLoss, Guid("CB2F6722-AB3A-11D2-9C40-00C04FA30A3E"), InterfaceType((short)1)]
public interface ICorRuntimeHost
{
    void CreateLogicalThreadState();
    void DeleteLogicalThreadState();
    void SwitchInLogicalThreadState([In] ref uint pFiberCookie);
    void SwitchOutLogicalThreadState([Out] IntPtr pFiberCookie);
    void LocksHeldByLogicalThread(out uint pCount);
    void MapFile([In] IntPtr hFile, out IntPtr hMapAddress);
    void GetConfiguration([MarshalAs(UnmanagedType.Interface)] out ICorConfiguration pConfiguration);
    void Start();
    void Stop();
    void CreateDomain([In, MarshalAs(UnmanagedType.LPWStr)] string pwzFriendlyName, [In, MarshalAs(UnmanagedType.IUnknown)] object pIdentityArray, [MarshalAs(UnmanagedType.IUnknown)] out object pAppDomain);
    void GetDefaultDomain([MarshalAs(UnmanagedType.IUnknown)] out object pAppDomain);
    void EnumDomains(out IntPtr hEnum);
    void NextDomain([In] IntPtr hEnum, [MarshalAs(UnmanagedType.IUnknown)] out object pAppDomain);
    void CloseEnum([In] IntPtr hEnum);
    void CreateDomainEx([In, MarshalAs(UnmanagedType.LPWStr)] string pwzFriendlyName, [In, MarshalAs(UnmanagedType.IUnknown)] object pSetup, [In, MarshalAs(UnmanagedType.IUnknown)] object pEvidence, [MarshalAs(UnmanagedType.IUnknown)] out object pAppDomain);
    void CreateDomainSetup([MarshalAs(UnmanagedType.IUnknown)] out object pAppDomainSetup);
    void CreateEvidence([MarshalAs(UnmanagedType.IUnknown)] out object pEvidence);
    void UnloadDomain([In, MarshalAs(UnmanagedType.IUnknown)] object pAppDomain);
    void CurrentDomain([MarshalAs(UnmanagedType.IUnknown)] out object pAppDomain);
}

그다음, ICorRuntimeHost 인터페이스를 구현한 객체를 구하고,

static void Main(string[] args)
{
    ICorRuntimeHost host = CoCreateInstance("{CB2F6723-AB3A-11D2-9C40-00C04FA30A3E}") as ICorRuntimeHost;
}

public static object CoCreateInstance(string guid)
{
    Guid coClassGuid = new Guid(guid);
    System.Type type = System.Type.GetTypeFromCLSID(coClassGuid);

    return Activator.CreateInstance(type);
} 

ICorRuntimeHost 객체로부터 IGCHost를 QI(QueryInterface)할 수 있습니다.

IGCHost gcHost32 = host as IGCHost;

이후, 단순하게 IGCHost.GetStats 메서드를 호출해 주는 것만으로 성능 카운터에서 본 해당 수치들을 가져올 수 있습니다.

public uint Flags;
public uint ExplicitGCCount;
public uint[] GenCollectionsTaken;
public uint CommittedKBytes;
public uint ReservedKBytes;
public uint Gen0HeapSizeKBytes;
public uint Gen1HeapSizeKBytes;
public uint Gen2HeapSizeKBytes;
public uint LargeObjectHeapSizeKBytes;
public uint KBytesPromotedFromGen0;
public uint KBytesPromotedFromGen1;

이렇게!

static void Main(string[] args)
{
    // ...[생략]...
    while (true)
    {
        long committedMBytes = GetGCState32(gcDetails, gcHost32);

        // ...[생략]...
    }
}

private static long GetGCState32(float[] gcDetails, IGCHost gcHost)
{
    _COR_GC_STATS stats = new _COR_GC_STATS();

    if (gcHost != null)
    {
        try
        {
            stats.Flags = (int)(COR_GC_STAT_TYPES.COR_GC_COUNTS | COR_GC_STAT_TYPES.COR_GC_MEMORYUSAGE);
            gcHost.GetStats(ref stats);

            gcDetails[0] = stats.Gen0HeapSizeKBytes;
            gcDetails[1] = stats.Gen1HeapSizeKBytes;
            gcDetails[2] = stats.Gen2HeapSizeKBytes;
            gcDetails[3] = stats.LargeObjectHeapSizeKBytes;

            return stats.CommittedKBytes;
        }
        catch
        {
        }
    }

    return 0;
}

주의할 점이 하나 있다면, IGCHost.GetStats 메서드의 호출이 응용 프로그램 초기에는 아무런 값도 반환하지 않는다는 것입니다. 그리고 32비트와 64비트에 따라 COR_GC_STATS의 구조체 내부 필드 크기가 달라진다는 것도 유의해야 하고.

(첨부한 프로젝트의 코드는 32비트/64비트에 따른 지원을 포함하고 있으니 참고하세요. ^^)




[이 글에 대해서 여러분들과 의견을 공유하고 싶습니다. 틀리거나 미흡한 부분 또는 의문 사항이 있으시면 언제든 댓글 남겨주십시오.]







[최초 등록일: ]
[최종 수정일: 12/16/2022]

Creative Commons License
이 저작물은 크리에이티브 커먼즈 코리아 저작자표시-비영리-변경금지 2.0 대한민국 라이센스에 따라 이용하실 수 있습니다.
by SeongTae Jeong, mailto:techsharer at outlook.com

비밀번호

댓글 작성자
 




1  [2]  3  4  5  6  7  8  9  10  11  12  13  14  15  ...
NoWriterDateCnt.TitleFile(s)
13893정성태2/27/20252211Linux: 115. eBPF (bpf2go) - ARRAY / HASH map 기본 사용법
13892정성태2/24/20252958닷넷: 2325. C# - PowerShell과 연동하는 방법파일 다운로드1
13891정성태2/23/20252489닷넷: 2324. C# - 프로세스의 성능 카운터용 인스턴스 이름을 구하는 방법파일 다운로드1
13890정성태2/21/20252305닷넷: 2323. C# - 프로세스 메모리 중 Private Working Set 크기를 구하는 방법(Win32 API)파일 다운로드1
13889정성태2/20/20253038닷넷: 2322. C# - 프로세스 메모리 중 Private Working Set 크기를 구하는 방법(성능 카운터, WMI) [1]파일 다운로드1
13888정성태2/17/20252483닷넷: 2321. Blazor에서 발생할 수 있는 async void 메서드의 부작용
13887정성태2/17/20253061닷넷: 2320. Blazor의 razor 페이지에서 code-behind 파일로 코드를 분리 및 DI 사용법
13886정성태2/15/20252572VS.NET IDE: 196. Visual Studio - Code-behind처럼 cs 파일을 그룹핑하는 방법
13885정성태2/14/20253229닷넷: 2319. ASP.NET Core Web API / Razor 페이지에서 발생할 수 있는 async void 메서드의 부작용
13884정성태2/13/20253501닷넷: 2318. C# - (async Task가 아닌) async void 사용 시의 부작용파일 다운로드1
13883정성태2/12/20253243닷넷: 2317. C# - Memory Mapped I/O를 이용한 PCI Configuration Space 정보 열람파일 다운로드1
13882정성태2/10/20252564스크립트: 70. 파이썬 - oracledb 패키지 연동 시 Thin / Thick 모드
13881정성태2/7/20252815닷넷: 2316. C# - Port I/O를 이용한 PCI Configuration Space 정보 열람파일 다운로드1
13880정성태2/5/20253155오류 유형: 947. sshd - Failed to start OpenSSH server daemon.
13879정성태2/5/20253377오류 유형: 946. Ubuntu - N: Updating from such a repository can't be done securely, and is therefore disabled by default.
13878정성태2/3/20253182오류 유형: 945. Windows - 최대 절전 모드 시 DRIVER_POWER_STATE_FAILURE 발생 (pacer.sys)
13877정성태1/25/20253236닷넷: 2315. C# - PCI 장치 열거 (레지스트리, SetupAPI)파일 다운로드1
13876정성태1/25/20253691닷넷: 2314. C# - ProcessStartInfo 타입의 Arguments와 ArgumentList파일 다운로드1
13875정성태1/24/20253123스크립트: 69. 파이썬 - multiprocessing 패키지의 spawn 모드로 동작하는 uvicorn의 workers
13874정성태1/24/20253541스크립트: 68. 파이썬 - multiprocessing Pool의 기본 프로세스 시작 모드(spawn, fork)
13873정성태1/23/20252968디버깅 기술: 217. WinDbg - PCI 장치 열거파일 다운로드1
13872정성태1/23/20252869오류 유형: 944. WinDbg - 원격 커널 디버깅이 연결은 되지만 Break (Ctrl + Break) 키를 눌러도 멈추지 않는 현상
13871정성태1/22/20253280Windows: 278. Windows - 윈도우를 다른 모니터 화면으로 이동시키는 단축키 (Window + Shift + 화살표)
13870정성태1/18/20253718개발 환경 구성: 741. WinDbg - 네트워크 커널 디버깅이 가능한 NIC 카드 지원 확대
13869정성태1/18/20253444개발 환경 구성: 740. WinDbg - _NT_SYMBOL_PATH 환경 변수에 설정한 경로로 심벌 파일을 다운로드하지 않는 경우
13868정성태1/17/20253096Windows: 277. Hyper-V - Windows 11 VM의 Enhanced Session 모드로 로그인을 할 수 없는 문제
1  [2]  3  4  5  6  7  8  9  10  11  12  13  14  15  ...