|
| 1 | +# SONiC Telemetry and Monitoring (TAM) High Level Design |
| 2 | + |
| 3 | +### Rev 0.1 |
| 4 | + |
| 5 | +## Table of Contents |
| 6 | + |
| 7 | +## 1. Revision |
| 8 | +Rev | Date | Author | Change Description |
| 9 | +----|------|--------|------------------- |
| 10 | +|v0.1|2025-10-09|Senthil Krishnamurthy|Initial version of TAM HLD |
| 11 | + |
| 12 | +## 2. Scope |
| 13 | +This document describes the high level design of Telemetry and Monitoring (TAM) in SONiC. |
| 14 | + |
| 15 | +## 3. Definitions/Abbreviations |
| 16 | +Definitions/Abbreviation|Description |
| 17 | +------------------------|----------- |
| 18 | +TAM| Telemetry and Monitoring |
| 19 | +SAI| Switch Abstraction Interface |
| 20 | +IPFIX| IP Flow Information Export |
| 21 | +VRF| Virtual Routing and Forwarding |
| 22 | +INT| In-band Network Telemetry |
| 23 | +IFA| In-band Flow Analyzer |
| 24 | + |
| 25 | +## 4. Overview |
| 26 | +TAM provides a framework to observe and report network events and telemetry by exporting messages to a collector. Telemetry provides information about the network traffic, such as congestion, latency, and port buffer utlization, on a per-device basis. Monitoring provides insights packet drops, and flow statistics. |
| 27 | + |
| 28 | +### 4.1 Telemetry |
| 29 | +Inband Telemetry (INT) is a variant of telemetry that provides insights into the path taken by the packets through the network, including per-hop latency, queue depth, and other metrics. Inband Flow Analyzer (IFA) is a mechanism INT uses to embed the telemetry data directly into the packet. In an IFA-aware network, the hops can perform one of the three roles: |
| 30 | +1. **Initiator Node**: This node initiates the IFA packet and can operate in one of two modes: |
| 31 | + a. **Inband Mode**: IFA header and the first IFA metadata, with telemetry data, is added to the data packet. |
| 32 | + b. **Probe Mode**: Incoming data packets are sampled and the first IFA metadata is added to the sampled packet. |
| 33 | +2. **Transit Node**: Matches on IFA packet and adds new IFA metadata to the packet. |
| 34 | +3. **Terminator Node**: Inband IFA packet are stripped off the IFA header and metadata, and the datapacket is forwarded to the next hop, and the metadata is exported to a collector. Probe IFA packets are teminated and mirroed to a collector. |
| 35 | + |
| 36 | +### 4.2 Monitoring |
| 37 | +TAM supports the Drop-Monitor (DM) feature, which provides visibility into packet drops in the dataplane. The dropped packets are trapped to a collector along with metadata that includes the reason for the drop and the payload of the packet, upto a maximum length. It can operate in two modes: |
| 38 | +1. **Stateless DM**: Drop reasons are exported on a per-packet per-drop basis. |
| 39 | +2. **Stateful DM**: Drop reasons are exported on a per-flow basis, in periodic intervals. |
| 40 | + |
| 41 | +## 5. Requirements |
| 42 | +TAM will be implemented in multiple phases. |
| 43 | + |
| 44 | +### 5.1 Phase 1 |
| 45 | +- Support Initiator and Transit roles for INT/IFA |
| 46 | +- Support IFA version 2.0 for the IFA header and metadata inserted in the packet |
| 47 | +- Support drop-monitor (DM) telemetry type with IPFIX reporting |
| 48 | +- Support stateless drop-monitoring |
| 49 | +- Collector endpoint can be either IPv4 or IPv6 |
| 50 | +- Collectors can be reachable via default or mgmt VRF (mgmt VRF must be enabled to use it) |
| 51 | +- Flow groups can be defined to seletcs packets for IFA and DM |
| 52 | +- TAM Session can be bound to one or more flow-groups |
| 53 | +- TAM Session can be bound to one or more collectors |
| 54 | + |
| 55 | +### 5.2 Phase 2 |
| 56 | +- Support stateful drop-monitoring |
| 57 | +- Support Terminator role for INT/IFA |
| 58 | +- Support packet sampling |
| 59 | + |
| 60 | +### 5.3 Phase 3 |
| 61 | +- Support additional telemetry types (e.g., flow-statistics) |
| 62 | +- Support additional report types (e.g., gRPC) |
| 63 | + |
| 64 | +## 6. Module Design |
| 65 | +### 6.1 Overall design |
| 66 | +- Management framework writes TAM tables to CONFIG_DB using yang model |
| 67 | +- TamOrch (new) in orchagent subscribes to CONFIG_DB TAM_* tables and programs SAI TAM objects |
| 68 | +- Syncd/SAI implement TAM object model and programs the dataplane |
| 69 | +- Telemetry data is exported to collectors from front-panel ports without punts to CPU |
| 70 | + |
| 71 | +### 6.2 Configuration and control flow |
| 72 | +The SWSS container is enhanced to add a new component, TamOrch, to process TAM configuration and control. |
| 73 | + |
| 74 | +#### 6.2.1 Inband Telemetry |
| 75 | +The following figure shows the configuration and control flows for TAM INT using IFAv2: |
| 76 | + |
| 77 | + |
| 78 | +1) Administrator configures TAM device attributes (CONFIG_DB), with IFA enabled |
| 79 | +2) tamorch uses aclorch to create an ACL table and an ACL rule to match on IP_PROTO 253 (IFA) |
| 80 | +3) tamorch creates SAI TAM_REPORT and TAM_INT objects |
| 81 | +4) The ACL table is bound to the switch |
| 82 | + |
| 83 | +#### 6.2.2 Drop Monitoring |
| 84 | +The following figure shows the configuration and control flows for TAM Drop Monitoring: |
| 85 | + |
| 86 | + |
| 87 | +1) Administrator configures TAM collectors, flow-groups, and sessions (CONFIG_DB) |
| 88 | +2) If the collector config uses a VRF, tamorch resolves the VRF to resolve the nexthop for the collector destination IP |
| 89 | +3) tamorch creates SAI TAM_COLLECTOR and TAM_TRANSPORT objects using the resolved nexthop |
| 90 | +4) tamorch uses aclorch to create an ACL table and an ACL rule with match conditions from the flow-group rules |
| 91 | +5) The ACL table is bound to the list of front-panel ports specified in the flow-group |
| 92 | +6) tamorch processes the TAM_SESSION_TABLE to create SAI TAM objects (report, event_action, event, tam) |
| 93 | + |
| 94 | +### 6.3 SWSS and syncd changes |
| 95 | +- New tamorch: consumes CONFIG_DB TAM tables, maps to SAI TAM objects, maintains reference counts and object lifecycles |
| 96 | +- syncd/SAI: no changes |
| 97 | + |
| 98 | +## 8. Configuration and Management |
| 99 | +### 8.1 CONFIG_DB |
| 100 | +Configure IFA by creating a TAM table: |
| 101 | +``` |
| 102 | +"TAM": { |
| 103 | + "device": { |
| 104 | + "device-id": 1234, // 28bits |
| 105 | + "enterprise-id": 1234, |
| 106 | + "ifa": true // boolean |
| 107 | + } |
| 108 | +} |
| 109 | +``` |
| 110 | + |
| 111 | +Configure drop-monitor by creating flow-groups, collectors, and sessions: |
| 112 | +``` |
| 113 | +"TAM_FLOW_GROUP": { |
| 114 | + "fg-1": { |
| 115 | + "aging_interval": 60, |
| 116 | + "ports": ["Ethernet0", "PortChannel10"] |
| 117 | + }, |
| 118 | + "fg-1|rule1": { |
| 119 | + "src_ip_prefix": "0.0.0.0/0", |
| 120 | + "dst_ip_prefix": "10.0.0.0/8", |
| 121 | + "ip_protocol": 6, |
| 122 | + "l4_dst_port": 443 |
| 123 | + } |
| 124 | +}, |
| 125 | +"TAM_COLLECTOR": { |
| 126 | + "c1": { |
| 127 | + "dst_ip": "192.0.2.10", |
| 128 | + "dst_port": 4739, |
| 129 | + "dscp_value": 32, |
| 130 | + "vrf": "vrf_blue" |
| 131 | + } |
| 132 | +}, |
| 133 | +"TAM_SESSION": { |
| 134 | + "s-drop": { |
| 135 | + "type": "drop-monitor", |
| 136 | + "report_type": "ipfix", |
| 137 | + "flow_group": "fg-1", |
| 138 | + "collector": ["c1"] |
| 139 | + } |
| 140 | +} |
| 141 | +``` |
| 142 | +Configure sFlow TAM Session: |
| 143 | +``` |
| 144 | +"TAM_SESSION": { |
| 145 | + "s-sflow": { |
| 146 | + "type": "sflow", |
| 147 | + "report_type": "ipfix", |
| 148 | + "collector": ["c1"] |
| 149 | + } |
| 150 | +} |
| 151 | +``` |
| 152 | + |
| 153 | +### 8.2 DB and Schema changes |
| 154 | + |
| 155 | +``` |
| 156 | +; Defines schema for TAM device configuration attributes |
| 157 | +key = TAM:device ; TAM device-level configuration |
| 158 | +; field = value |
| 159 | +DEVICE_ID = 1*9DIGIT ; 1..134217727 |
| 160 | +ENTERPRISE_ID = 1*9DIGIT ; 1..134217727 |
| 161 | +IFA = "true" / "false" ; Enable IFA device-type hint |
| 162 | +``` |
| 163 | + |
| 164 | +``` |
| 165 | +; Defines schema for TAM flow-group configuration attributes |
| 166 | +key = TAM_FLOW_GROUP:flow_group_name |
| 167 | +; field = value |
| 168 | +AGING_INTERVAL = 1*10DIGIT ; interval in seconds |
| 169 | +PORTS = ifname-list ; comma-separated list of interfaces |
| 170 | +
|
| 171 | +; value annotations |
| 172 | +flow_group_name = 1*255VCHAR |
| 173 | +ifname-list = ifname *( "," ifname ) |
| 174 | +ifname = 1*64VCHAR |
| 175 | +``` |
| 176 | + |
| 177 | +``` |
| 178 | +; Defines schema for TAM flow-group match rules |
| 179 | +key = TAM_FLOW_GROUP:flow_group_name|rule_name |
| 180 | +; field = value |
| 181 | +SRC_IP_PREFIX = ip_prefix ; mandatory |
| 182 | +DST_IP_PREFIX = ip_prefix ; mandatory |
| 183 | +L4_SRC_PORT = port_num ; optional |
| 184 | +L4_DST_PORT = port_num ; optional |
| 185 | +IP_PROTOCOL = 1*3DIGIT ; 1..255 (optional) |
| 186 | +
|
| 187 | +; value annotations |
| 188 | +rule_name = 1*64VCHAR |
| 189 | +ip_prefix = IPv4prefix / IPv6prefix |
| 190 | +IPv4prefix = IPv4address "/" 1*2DIGIT ; 0..32 |
| 191 | +IPv6prefix = IPv6address "/" 1*3DIGIT ; 0..128 |
| 192 | +port_num = 1*5DIGIT ; 1..65535 |
| 193 | +``` |
| 194 | + |
| 195 | +``` |
| 196 | +; Defines schema for TAM collector configuration attributes |
| 197 | +key = TAM_COLLECTOR:collector_name |
| 198 | +; field = value |
| 199 | +SRC_IP = IPv4address / IPv6address ; optional |
| 200 | +DST_IP = IPv4address / IPv6address ; mandatory |
| 201 | +DST_PORT = port_num ; mandatory |
| 202 | +DSCP_VALUE = dscp ; DSCP 0..63 (default 32) |
| 203 | +VRF = "default" / "mgmt" / vrf_name ; VRF for collector reachability |
| 204 | +
|
| 205 | +; value annotations |
| 206 | +collector_name = 1*255VCHAR |
| 207 | +vrf_name = 1*255VCHAR |
| 208 | +dscp = 1*2DIGIT ; 0..63 |
| 209 | +``` |
| 210 | + |
| 211 | +``` |
| 212 | +; Defines schema for TAM session configuration attributes |
| 213 | +key = TAM_SESSION:session_name |
| 214 | +; field = value |
| 215 | +TYPE = "drop-monitor" |
| 216 | +REPORT_TYPE = "ipfix" |
| 217 | +FLOW_GROUP = flow_group_name |
| 218 | +COLLECTOR = collector_list ; one or more collectors |
| 219 | +
|
| 220 | +; value annotations |
| 221 | +session_name = 1*255VCHAR |
| 222 | +collector_list = collector_name *( "," collector_name ) |
| 223 | +``` |
| 224 | + |
| 225 | +> Note: Refer to swss-schema.md for standard value annotations such as IPv4address/IPv6address and ifname, and for general BNF conventions used across SONiC documents. |
| 226 | +
|
| 227 | + |
| 228 | +## 9. Warmboot and Fastboot Impact |
| 229 | +- No additional sleeps in boot-critical path. TAM object creation occurs after dependencies are up. Service can be delayed until SYSTEM_READY is Up. When disabled/unused, no impact. |
| 230 | + |
| 231 | +## 10. Memory Consumption |
| 232 | +- Minimal control-plane state in orchagent (object maps). No growth when feature disabled. |
| 233 | + |
| 234 | +## 11. Restrictions/Limitations |
| 235 | +- Requires platform/SAI support for TAM drop monitoring and IPFIX export; otherwise feature remains inoperative (capability=false) |
| 236 | +- Exact limits (number of flow-groups/rules/collectors) depend on platform |
| 237 | +- mgmt VRF usage requires MGMT_VRF enabled |
| 238 | + |
| 239 | +## 12. Testing Requirements |
| 240 | +### 12.1 Unit tests (one-liners) |
| 241 | +1) Validate CONFIG_DB for each table and field |
| 242 | +2) Validate reference checks (ports, VRF, flow_group, collector) |
| 243 | +3) Validate tamorch creates/updates/deletes SAI TAM objects per CONFIG_DB changes |
| 244 | +4) Capability gating: with capability=false, CONFIG_DB writes do not program SAI |
| 245 | + |
| 246 | +### 12.2 System tests |
| 247 | +1) Configure flow-group, rule, collector, session; verify IPFIX is exported to collector |
| 248 | +2) Verify VRF selection (default vs mgmt) and DSCP marking |
| 249 | +3) Verify rule match scoping and port/PortChannel membership |
| 250 | +4) Reboot/warm-reboot and verify export resumes with preserved configuration |
| 251 | + |
| 252 | +## 13. Open/Action items |
| 253 | +- Finalize CLI commands and Command-Reference.md updates aligned to YANG |
| 254 | +- Document per-platform capability and limits |
0 commit comments