5G AMF过载控制：N2节流与NAS退避信令全链路拆解

AMF内部过载检测度量:
├── CPU利用率（per-AMF实例）
├── 内存使用率（UE Context数量 vs 容量上限）
├── NAS消息队列深度（待处理的Initial UE Message数量）
├── N2信令处理速率（NGAP消息/秒 vs 处理能力上限）
├── 数据库访问延迟（UDM/UDR查询响应时间）
└── 线程池/进程池饱和度

sequenceDiagram
    participant UE as UE
    participant gNB as gNB(NG-RAN)
    participant AMF as AMF
    participant PCF as PCF
    participant UDM as UDM

    Note over AMF: AMF检测到过载<br/>(CPU/内存/队列超阈值)

    rect rgb(255, 230, 230)
    Note over AMF,gNB: === 过载控制启动阶段 ===

    AMF->>gNB: NGAP OVERLOAD START<br/>+ Overload Response IE<br/>+ Traffic Load Reduction Indication
    Note right of AMF: 指示gNB执行流量抑制策略
    Note right of gNB: gNB根据Overload Response<br/>执行差异化流量控制
    end

    rect rgb(255, 255, 220)
    Note over UE,gNB: === 过载期间的信令处理 ===

    UE->>gNB: RRC Setup Request<br/>(非紧急MO信令)
    gNB-->>UE: RRC Reject<br/>+ waitTime
    Note right of gNB: 根据Overload Response<br/>拒绝非紧急MO信令

    UE->>gNB: RRC Setup Request<br/>(Emergency/High Priority)
    gNB->>AMF: NGAP Initial UE Message<br/>(允许紧急/高优先级通过)
    AMF->>gNB: NGAP Initial Context Setup Request
    gNB->>UE: RRC Reconfiguration
    Note right of gNB: 紧急呼叫和高优先级<br/>始终放行

    UE->>gNB: RRC Setup Request<br/>(普通MO Data)
    gNB-->>UE: RRC Reject<br/>+ waitTime
    Note right of gNB: 根据策略拒绝/延迟<br/>普通数据连接请求
    end

    rect rgb(220, 255, 220)
    Note over AMF,gNB: === NAS层协同退避 ===

    UE->>gNB: RRC Setup Request<br/>(MT Data - 被允许)
    gNB->>AMF: NGAP Initial UE Message<br/>(Service Request)
    AMF-->>UE: NAS Registration/Service Reject<br/>+ Cause: Congestion<br/>+ T3346 Backoff Timer
    Note right of AMF: AMF也通过NAS层<br/>直接拒绝部分请求
    end

    rect rgb(220, 220, 255)
    Note over AMF,gNB: === 过载恢复阶段 ===

    Note over AMF: AMF检测到负载降低<br/>低于恢复阈值

    AMF->>gNB: NGAP OVERLOAD STOP
    Note right of gNB: gNB恢复正常操作<br/>接受所有RRC连接请求

    UE->>gNB: RRC Setup Request<br/>(之前被拒绝的UE重试)
    gNB->>AMF: NGAP Initial UE Message
    Note right of gNB: 正常处理流程恢复
    end

OVERLOAD START ::= SEQUENCE {
    protocolIEs    ProtocolIE-Container { {OverloadStartIEs} },
    ...
}

OverloadStartIEs NGAP-ELEMENTARY-PROCEDURE ::= {
    { ID id-OverloadResponse    CRITICALITY reject    TYPE OverloadResponse    PRESENCE mandatory } |
    { ID id-TrafficLoadReductionIndication   CRITICALITY ignore   TYPE TrafficLoadReductionIndication   PRESENCE optional } |
    { ID id-OverloadStartNSSAIList           CRITICALITY ignore   TYPE OverloadStartNSSAIList           PRESENCE optional },
    ...
}

gNB流量抑制实现逻辑（伪代码）:
if (random(0, 100) < trafficLoadReductionIndication) {
    reject RRC Connection Request;
    send RRC Reject with waitTime;
} else {
    allow RRC Connection Request;
    forward to AMF via NGAP Initial UE Message;
}

OverloadStartNSSAIList ::= SEQUENCE (SIZE(1..maxnoofSliceItems)) OF OverloadStartNSSAIItem

OverloadStartNSSAIItem ::= SEQUENCE {
    sliceOverloadResponse          SliceOverloadResponse,
    sliceTrafficLoadReductionIndication  SliceTrafficLoadReductionIndication OPTIONAL,
    s-NSSAI                         S-NSSAI
}

OVERLOAD STOP ::= SEQUENCE {
    protocolIEs    ProtocolIE-Container { {OverloadStopIEs} }
}

flowchart TD
    A[收到OVERLOAD START] --> B{解析Overload Response IE}
    B --> C[存储过载策略到gNB上下文]
    C --> D{收到RRC Connection Request}
    D --> E{检查UE Access Category<br/>和Access Identity}
    E --> F{UE是否属于<br/>豁免类别?}
    F -->|Emergency| G[允许RRC建立<br/>正常转发至AMF]
    F -->|High Priority AI 11-15| G
    F -->|MT Service<br/>寻呼响应| H{Overload Response<br/>是否允许MT?}
    H -->|permit-emergency-mt| G
    H -->|permit-emergency-mt-hp| G
    H -->|reject-all-signalling| I[拒绝RRC建立]
    F -->|普通MO信令/数据| J{应用Traffic Load<br/>Reduction概率}
    J -->|random < threshold| I
    J -->|random >= threshold| G
    I --> K[发送RRC Reject<br/>+ waitTime]

    L[收到OVERLOAD STOP] --> M[清除过载策略]
    M --> N[恢复正常RRC接纳逻辑]

flowchart TD
    subgraph 第一层防线
        A[gNB] -->|Overload Start| B[拒绝部分RRC请求<br/>源头压制]
    end

    subgraph 第二层防线
        C[AMF] -->|NAS Reject + T3346| D[已到达AMF的请求<br/>直接退避]
    end

    B -->|漏过的流量| C
    D --> E[UE进入退避状态<br/>不再发起请求]

GPRS Timer 3 IE结构:
┌────────────────────────────────────────┐
│  Bit 8  7  6  5  4  3  2  1           │
├────────────────────────────────────────┤
│ Byte 1: [Unit][Timer Value (5 bits)]   │
│   Unit: 00=2秒, 01=1分, 10=6分, 11=拒绝│
│   Timer Value: 0-31 (乘以Unit)         │
├────────────────────────────────────────┤
│ Byte 2: 扩展值 (如果使用扩展格式)      │
└────────────────────────────────────────┘

示例: Unit=01(1分钟), Value=30 → T3346=30分钟

flowchart TB
    subgraph AMF过载检测
        A[AMF检测到过载]
    end

    subgraph 切片级策略
        B[S-NSSAI: IoT切片<br/>Traffic Reduction: 80%<br/>Response: reject-all-signalling]
        C[S-NSSAI: eMBB切片<br/>Traffic Reduction: 20%<br/>Response: reject-non-emergency-mo-data]
        D[S-NSSAI: URLLC切片<br/>不包含在Overload Start中<br/>→ 不受影响]
    end

    A --> B
    A --> C
    A --> D

    subgraph gNB差异化执行
        E[IoT UE → 80%概率拒绝]
        F[eMBB UE → 20%概率拒绝]
        G[URLLC UE → 正常服务]
    end

    B --> E
    C --> F
    D --> G

flowchart LR
    subgraph AMF Set
        A1[AMF-1<br/>Weight: 0.1<br/>过载]
        A2[AMF-2<br/>Weight: 0.5<br/>正常]
        A3[AMF-3<br/>Weight: 0.4<br/>正常]
    end

    subgraph gNB
        G1[新UE接入]
    end

    G1 -->|10%概率| A1
    G1 -->|50%概率| A2
    G1 -->|40%概率| A3

    A1 -.->|Overload Start| G1
    A1 -.->|Config Update<br/>Weight: 0→0.1| G1

# Wireshark/tshark过滤NGAP Overload Start
ngap.procedureCode == 17

# 过滤Overload Stop
ngap.procedureCode == 18

# 过滤携带T3346的NAS Reject
nas_5gs.message_type == 0x44 && nas_5gs.cause == 22

NGAP Overload Start
├── AMF Overload Response
│   └── reject-non-emergency-mo-data (1)
├── Traffic Load Reduction Indication
│   └── 50 (即50%流量抑制)
└── Overload Start NSSAI List (如果存在)
    ├── Item 1:
    │   ├── S-NSSAI: SST=1, SD=010203
    │   ├── Slice Overload Response: reject-all-signalling
    │   └── Slice Traffic Load Reduction: 80
    └── Item 2:
        ├── S-NSSAI: SST=2, SD=040506
        ├── Slice Overload Response: reject-non-emergency-mo-data
        └── Slice Traffic Load Reduction: 30

stateDiagram-v2
    [*] --> Normal: AMF启动
    Normal --> Overload_Level1: 度量超过阈值L1
    Normal --> Overload_Level2: 度量超过阈值L2
    Normal --> Overload_Level3: 度量超过阈值L3

    Overload_Level1 --> Overload_Level2: 度量恶化超L2阈值
    Overload_Level1 --> Normal: 度量低于恢复阈值<br/>发送OVERLOAD STOP
    Overload_Level2 --> Overload_Level3: 度量恶化超L3阈值
    Overload_Level2 --> Overload_Level1: 度量改善至L1范围
    Overload_Level3 --> Overload_Level2: 度量改善至L2范围

    Overload_Level1: 轻度过载<br/>N2 Overload Start<br/>Response: permit-emergency-mt-hp
    Overload_Level2: 中度过载<br/>升级Overload Response<br/>+ NAS Reject部分请求
    Overload_Level3: 严重过载<br/>最严格Response<br/>+ 全量NAS Reject + T3346
    Normal: 正常运行<br/>无过载控制约束

    note right of Overload_Level1
        发送NGAP OVERLOAD START
        Traffic Reduction: 20-40%
    end note

    note right of Overload_Level3
        发送升级的OVERLOAD START
        Traffic Reduction: 60-100%
        可能触发AMF重平衡
    end note