Zack Williams | 553a363 | 2019-08-09 17:14:43 -0700 | [diff] [blame] | 1 | Architecture and Design |
| 2 | *********************** |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 3 | |
Charles Chan | 6613eac | 2019-09-17 15:42:48 -0700 | [diff] [blame] | 4 | Overview |
| 5 | -------- |
| 6 | |
| 7 | .. image:: images/arch-overview.png |
| 8 | :width: 1000px |
| 9 | |
| 10 | Trellis operates as a **hybrid L2/L3 fabric**. |
| 11 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 12 | As a **pure (or classic) SDN** solution, Trellis **does not use any of the |
| 13 | traditional control protocols** typically found in networking, a non-exhaustive |
| 14 | list of which includes: STP, MSTP, RSTP, LACP, MLAG, PIM, IGMP, OSPF, IS-IS, |
| 15 | Trill, RSVP, LDP and BGP. |
Charles Chan | 6613eac | 2019-09-17 15:42:48 -0700 | [diff] [blame] | 16 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 17 | Instead, Trellis **uses an SDN Controller (ONOS) decoupled from the data-plane |
| 18 | hardware to directly program ASIC forwarding tables using OpenFlow and with |
| 19 | OF-DPA**, an open-API from Broadcom running on the switches. |
Charles Chan | 6613eac | 2019-09-17 15:42:48 -0700 | [diff] [blame] | 20 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 21 | In this design, a set of applications running on ONOS implement all the fabric |
| 22 | functionality and features, such as **Ethernet switching**, **IP routing**, |
| 23 | **multicast**, **DHCP Relay**, **pseudowires** and more. |
Charles Chan | 6613eac | 2019-09-17 15:42:48 -0700 | [diff] [blame] | 24 | |
| 25 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 26 | You can learn more about Trellis features and design concepts by visiting |
| 27 | the `Project Website <https://opennetworking.org/trellis>`_ and reading the |
| 28 | `Platform Brief |
| 29 | <https://www.opennetworking.org/wp-content/uploads/2019/09/TrellisPlatformBrief.pdf>`_. |
Charles Chan | 6613eac | 2019-09-17 15:42:48 -0700 | [diff] [blame] | 30 | |
| 31 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 32 | Introduction to OF-DPA Pipeline |
| 33 | ------------------------------- |
| 34 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 35 | In this design note, we are going to explain the design choices we have made |
| 36 | and how we got around OF-DPA (OpenFlow Data Plane Abstraction) pipeline |
| 37 | restrictions to implement the features we need. We will start from explaining |
| 38 | the OFDPA flow tables and group tables we use. |
| 39 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 40 | Fig. 1 shows the simplified OFDPA pipeline overview. |
| 41 | |
| 42 | .. image:: images/arch-ofdpa.png |
| 43 | :width: 1000px |
| 44 | |
| 45 | Fig. 1 Simplified OF-DPA pipeline overview |
| 46 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 47 | Flow Tables |
| 48 | ----------- |
| 49 | |
| 50 | VLAN Table |
| 51 | ^^^^^^^^^^ |
| 52 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 53 | The **VLAN Flow Table (id=10)** is used for IEEE 801.Q VLAN assignment and |
| 54 | filtering to specify how VLANs are to be handled on a particular port. |
| 55 | **All packets must have an associated VLAN id in order to be processed by |
| 56 | subsequent tables**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 57 | |
| 58 | **Table miss**: goto **ACL table**. |
| 59 | |
| 60 | According to OFDPA spec, we need to assign a VLAN ID even for untagged packets. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 61 | Each untagged packet will be tagged with an **internal VLAN** when being |
| 62 | handled by VLAN table. The internal VLAN will be popped when the packet is |
| 63 | sent to a output port or controller. The internal VLAN is assigned according |
| 64 | to the subnet configuration of the input port. Packets coming from ports that |
| 65 | do not have subnet configured (e.g. the spine facing ports) will be tagged with |
| 66 | VLAN ID **4094**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 67 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 68 | The internal VLAN is also used to determine the subnet when a packet needs to |
| 69 | be flooded to all ports in the same subnet. (See L2 Broadcast section for |
| 70 | detail.) |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 71 | |
| 72 | Termination MAC Table |
| 73 | ^^^^^^^^^^^^^^^^^^^^^ |
| 74 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 75 | The **Termination MAC (TMAC) Flow Table (id=20)** determines whether to do |
| 76 | bridging or routing on a packet. |
| 77 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 78 | It identifies routed packets their destination MAC, VLAN, and Ethertype. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 79 | |
| 80 | Routed packet rule types use a Goto-Table instruction to indicate that the |
| 81 | next table is one of the routing tables. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 82 | |
| 83 | **Table miss**: goto **Bridging table**. |
| 84 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 85 | In this table, we determine which table the packet should go to by checking the |
| 86 | destination MAC address and the Ethernet type of the packet. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 87 | |
| 88 | - if dst_mac = router MAC and eth_type = ip, goto **unicast routing** table |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 89 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 90 | - if dst_mac = router MAC and eth_type = mpls, goto **MPLS table** |
| 91 | |
| 92 | - if dst_mac = multicast MAC (01:00:5F:00:00:00/FF:FF:FF:80:00:00), goto |
| 93 | **multicast routing** table |
| 94 | |
| 95 | - none of above, goto **bridging table** |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 96 | |
| 97 | MPLS Tables |
| 98 | ^^^^^^^^^^^ |
| 99 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 100 | The MPLS pipeline can support three **MPLS Flow Tables, MPLS Table 0 |
| 101 | (id=23), MPLS Table 1 (id=24) and MPLS Table 2 (id=25)**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 102 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 103 | An MPLS Flow Table lookup matches the label in the outermost MPLS shim |
| 104 | header in the packets. |
| 105 | |
| 106 | - MPLS Table 0 is only used to pop a protection label on platforms that |
| 107 | support this table, or to detect an MPLS- TP Section OAM PDU. |
| 108 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 109 | - MPLS Table 1 and MPLS Table 2 can be used for all label operations. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 110 | |
| 111 | - MPLS Table 1 and MPLS Table 2 are synchronized flow tables and updating |
| 112 | one updates the other |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 113 | |
| 114 | **Table miss**: goto **ACL table**. |
| 115 | |
| 116 | We only use MPLS Table 1 (id=24) in current design. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 117 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 118 | MPLS packets are matched by the MPLS label. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 119 | |
| 120 | The packet will go to **L3 interface group** with MPLS label being popped and |
| 121 | further go to destination leaf switch. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 122 | |
| 123 | |
| 124 | Unicast Routing Table |
| 125 | ^^^^^^^^^^^^^^^^^^^^^ |
| 126 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 127 | The **Unicast Routing Flow Table (id=30)** supports routing for potentially |
| 128 | large numbers of IPv4 and IPv6 flow entries using the hardware L3 tables. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 129 | |
| 130 | **Table miss**: goto **ACL table**. |
| 131 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 132 | In this table, we determine where to output a packet by checking its |
| 133 | **destination IP (unicast)** address. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 134 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 135 | - if dst_ip locates at a **remote switch**, the packet will go to an **L3 ECMP |
| 136 | group**, be tagged with MPLS label, and further go to a spine switch |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 137 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 138 | - if dst_ip locates at the **same switch**, the packet will go to an **L3 |
| 139 | interface group** and further go to a host |
| 140 | |
| 141 | Note that the priority of flow entries in this table is sorted by prefix |
| 142 | length. |
| 143 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 144 | Longer prefix (/32) will have higher priority than shorter prefix (/0). |
| 145 | |
| 146 | |
| 147 | Multicast Routing Table |
| 148 | ^^^^^^^^^^^^^^^^^^^^^^^ |
| 149 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 150 | The **Multicast Routing Flow Table (id=40)** supports routing for IPv4 and |
| 151 | IPv6 multicast packets. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 152 | |
| 153 | **Table miss**: goto **ACL table**. |
| 154 | |
| 155 | Flow entries in this table always match the **destination IP (multicast)**. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 156 | |
| 157 | Matched packets will go to an **L3 multicast group** and further go to the next |
| 158 | switch or host. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 159 | |
| 160 | |
| 161 | Bridging Table |
| 162 | ^^^^^^^^^^^^^^ |
| 163 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 164 | The **Bridging Flow Table (id=50)** supports Ethernet packet switching for |
| 165 | potentially large numbers of flow entries using the hardware L2 tables. |
| 166 | |
| 167 | The Bridging Flow Table forwards either based on VLAN (normal switched |
| 168 | packets) or Tunnel id (isolated forwarding domain packets), with the Tunnel |
| 169 | id metadata field used to distinguish different flow table entry types by |
| 170 | range assignment. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 171 | |
| 172 | **Table miss**: goto **ACL table**. |
| 173 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 174 | In this table, we match the **VLAN ID** and the **destination MAC address** and |
| 175 | determine where the packet should be forwarded to. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 176 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 177 | - if the destination MAC can be matched, the packet will go to the **L2 |
| 178 | interface group** and further sent to the destination host. |
| 179 | |
| 180 | - if the destination MAC can not be matched, the packet will go to the **L2 |
| 181 | flood group** and further flooded to the same subnet. |
| 182 | |
| 183 | Since we cannot match IP in bridging table, we use the VLAN ID to determine |
| 184 | which subnet this packet should be flooded to. |
| 185 | |
| 186 | The VLAN ID can be either (1) the internal VLAN assigned to untagged packets |
| 187 | in VLAN table or (2) the VLAN ID that comes with tagged packets. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 188 | |
| 189 | |
| 190 | Policy ACL Table |
| 191 | ^^^^^^^^^^^^^^^^ |
| 192 | .. note:: |
| 193 | The Policy ACL Flow Table supports wide, multi-field matching. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 194 | |
| 195 | Most fields can be wildcard matched, and relative priority must be |
| 196 | specified in all flow entry modification API calls. |
| 197 | |
| 198 | This is the preferred table for matching BPDU and ARP packets. It also |
| 199 | provides the Metering instruction. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 200 | |
| 201 | **Table miss**: **do nothing**. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 202 | The packet will be forwarded using the output or group in the action set, |
| 203 | if any. |
| 204 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 205 | If the action set does not have a group or output action the packet is dropped. |
| 206 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 207 | In ACL table we trap **ARP**, **LLDP**, **BDDP**, **DHCP** and send those |
| 208 | packets to the **controller**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 209 | |
| 210 | Group Tables |
| 211 | ------------ |
| 212 | |
| 213 | L3 ECMP Group |
| 214 | ^^^^^^^^^^^^^ |
| 215 | .. note:: |
| 216 | OF-DPA L3 ECMP group entries are of OpenFlow type **SELECT**. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 217 | |
| 218 | For IP routing the action buckets reference the OF-DPA **L3 Unicast group** |
| 219 | entries that are members of the multipath group for ECMP forwarding. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 220 | |
| 221 | An OF-DPA L3 ECMP Group entry can also be used in a Provider Edge Router. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 222 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 223 | In this packet flow it can chain to either an **MPLS L3 Label** group entry |
| 224 | or to an **MPLS Fast Failover** group entry. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 225 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 226 | An OF-DPA L3 ECMP Group entry can be specified as a routing target instead |
| 227 | of an OF-DPA L3 Unicast Group entry. Selection of an action bucket for |
| 228 | forwarding a particular packet is hardware-specific. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 229 | |
| 230 | MPLS Label Group |
| 231 | ^^^^^^^^^^^^^^^^ |
| 232 | .. note:: |
| 233 | MPLS Label Group entries are of OpenFlow **INDIRECT** type. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 234 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 235 | There are four MPLS label Group entry subtypes, all with similar structure. |
| 236 | |
| 237 | These can be used in different configurations to **push up to three |
| 238 | labels** for tunnel initiation or LSR swap. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 239 | |
| 240 | MPLS Interface Group |
| 241 | ^^^^^^^^^^^^^^^^^^^^ |
| 242 | .. note:: |
| 243 | MPLS Interface Group Entry is of OpenFlow type **INDIRECT**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 244 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 245 | It is used to **set the outgoing L2 header** to reach the next hop label |
| 246 | switch router or provider edge router. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 247 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 248 | We use **L3 ECMP** group to randomly pick one spine switch when we need to |
| 249 | route a packet from leaves to spines. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 250 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 251 | We point each bucket to an **MPLS Label** Group in which the MPLS labels are |
| 252 | pushed to the packets to realize Segment Routing mechanism. (More |
| 253 | specifically, we use the subtype 2 **MPLS L3 VPN Label**). |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 254 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 255 | We then point an MPLS Label Group points to an **MPLS Interface** Group in |
| 256 | which the destination MAC is set to the next hop (spine router). |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 257 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 258 | Finally, the packet will goto an **L2 Interface** Group and being sent to the |
| 259 | output port that goes to the spine router. |
| 260 | |
| 261 | Detail of how segment routing is implemented will be explained in the L3 |
| 262 | unicast section below. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 263 | |
| 264 | L3 Unicast Group |
| 265 | ^^^^^^^^^^^^^^^^ |
| 266 | .. note:: |
| 267 | OF-DPA L3 Unicast group entries are of OpenFlow **INDIRECT** type. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 268 | |
| 269 | L3 Unicast group entries are used to supply the routing next hop and output |
| 270 | interface for packet forwarding. |
| 271 | |
| 272 | To properly route a packet from either the Routing Flow Table or the Policy |
| 273 | ACL Flow Table, the forwarding flow entry must reference an L3 Unicast |
| 274 | Group entry. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 275 | |
| 276 | All packets must have a VLAN tag. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 277 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 278 | **A chained L2 Interface group entry must be in the same VLAN as assigned |
| 279 | by the L3 Unicast Group** entry. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 280 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 281 | We use L3 Unicast Group to rewrite the **source MAC**, **destination MAC** and |
| 282 | **VLAN ID** when routing is needed. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 283 | |
| 284 | L3 Multicast Group |
| 285 | ^^^^^^^^^^^^^^^^^^ |
| 286 | .. note:: |
| 287 | OF-DPA L3 Multicast group entries are of OpenFlow **ALL** type. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 288 | |
| 289 | The action buckets describe the interfaces to which multicast packet |
| 290 | replicas are forwarded. |
| 291 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 292 | Note that: |
| 293 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 294 | - Chained OF-DPA **L2 Interface** Group entries must be in the **same |
| 295 | VLAN** as the OF-DPA **L3 Multicast** group entry. However, |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 296 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 297 | - Chained OF-DPA **L3 Interface** Group entries must be in **different |
| 298 | VLANs** from the OF-DPA **L3 Multicast** Group entry, **and from each |
| 299 | other**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 300 | |
| 301 | We use L3 multicast group to replicate multicast packets when necessary. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 302 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 303 | It is also possible that L3 multicast group consists of only one bucket when |
| 304 | replication is not needed. |
| 305 | |
| 306 | Detail of how multicast is implemented will be explained in the L3 multicast |
| 307 | section below. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 308 | |
| 309 | L2 Interface Group |
| 310 | ^^^^^^^^^^^^^^^^^^ |
| 311 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 312 | L2 Interface Group entries are of OpenFlow **INDIRECT** type, with a single |
| 313 | action bucket. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 314 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 315 | OF-DPA L2 Interface group entries are used for egress VLAN filtering and |
| 316 | tagging. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 317 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 318 | If a specific set of VLANs is allowed on a port, appropriate group entries |
| 319 | must be defined for the VLAN and port combinations. |
| 320 | |
| 321 | Note: OF-DPA uses the L2 Interface group declaration to configure the port |
| 322 | VLAN filtering behavior. |
| 323 | |
| 324 | This approach was taken since OpenFlow does not support configuring VLANs |
| 325 | on physical ports. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 326 | |
| 327 | L2 Flood Group |
| 328 | ^^^^^^^^^^^^^^ |
| 329 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 330 | L2 Flood Group entries are used by VLAN Flow Table wildcard (destination |
| 331 | location forwarding, or DLF) rules. |
| 332 | |
| 333 | Like OF-DPA L2 Multicast group entry types they are of OpenFlow **ALL** |
| 334 | type. |
| 335 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 336 | The action buckets each encode an output port. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 337 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 338 | Each OF-DPA L2 Flood Group entry bucket forwards a replica to an output |
| 339 | port, except for packet IN_PORT. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 340 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 341 | All of the OF-DPA L2 Interface Group entries referenced by the OF-DPA Flood |
| 342 | Group entry, and the OF- DPA Flood Group entry itself, must be in the |
| 343 | **same VLAN**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 344 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 345 | Note: There can only be **one OF-DPA L2 Flood Group** entry defined **per |
| 346 | VLAN**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 347 | |
| 348 | L2 Unicast |
| 349 | ---------- |
| 350 | |
| 351 | .. image:: images/arch-l2u.png |
| 352 | :width: 800px |
| 353 | |
| 354 | Fig. 2: L2 unicast |
| 355 | |
| 356 | .. image:: images/arch-l2u-pipeline.png |
| 357 | :width: 1000px |
| 358 | |
| 359 | Fig. 3: Simplified L2 unicast pipeline |
| 360 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 361 | The L2 unicast mechanism is designed to support **intra-rack (intra-subnet)** |
| 362 | communication when the destination host is **known**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 363 | |
| 364 | Pipeline Walkthrough - L2 Unicast |
| 365 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 366 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 367 | - **VLAN Table**: An untagged packet will be assigned an internal VLAN ID |
| 368 | according to the input port and the subnet configured on the input port. |
| 369 | Packets of the same subnet will have the same internal VLAN ID. |
| 370 | |
| 371 | - **TMAC Table**: Since the destination MAC of a L2 unicast packet is not the |
| 372 | MAC of leaf router, the packet will miss the TMAC table and goes to the |
| 373 | bridging table. |
| 374 | |
| 375 | - **Bridging Table**: If the destination MAC is learnt, there will be a flow |
| 376 | entry matching that destination MAC and pointing to an L2 interface group. |
| 377 | |
| 378 | - **ACL Table**: IP packets will miss the ACL table and the L2 interface group |
| 379 | will be executed. |
| 380 | |
| 381 | L2 Interface Group: The internal assigned VLAN will be popped before the |
| 382 | packet is sent to the output port. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 383 | |
| 384 | L2 Broadcast |
| 385 | ------------ |
| 386 | |
| 387 | .. image:: images/arch-l2f.png |
| 388 | :width: 800px |
| 389 | |
| 390 | Fig. 4: L2 broadcast |
| 391 | |
| 392 | .. image:: images/arch-l2f-pipeline.png |
| 393 | :width: 1000px |
| 394 | |
| 395 | Fig. 5: Simplified L2 broadcast pipeline |
| 396 | |
| 397 | Pipeline Walkthrough - L2 Broadcast |
| 398 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 399 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 400 | - **VLAN Table**: (same as L2 unicast) |
| 401 | |
| 402 | - **TMAC Table**: (same as L2 unicast) |
| 403 | |
| 404 | - **Bridging Table**: If the destination MAC is not learnt, there will NOT be a |
| 405 | flow entry matching that destination MAC. |
| 406 | |
| 407 | It will then fallback to a lower priority entry that matches the VLAN |
| 408 | (subnet) and point to an L2 flood group. |
| 409 | |
| 410 | - **ACL Table**: IP packets will miss the ACL table and the L2 flood group will |
| 411 | be executed. |
| 412 | |
| 413 | - **L2 Flood Group**: Consists of all L2 interface groups related to this VLAN |
| 414 | (subnet). |
| 415 | |
| 416 | - **L2 Interface Group**: The internal assigned VLAN will be popped before the |
| 417 | packet is sent to the output port. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 418 | |
| 419 | ARP |
| 420 | --- |
| 421 | |
| 422 | .. image:: images/arch-arp-pipeline.png |
| 423 | :width: 1000px |
| 424 | |
| 425 | Fig. 6: Simplified ARP pipeline |
| 426 | |
| 427 | All ARP packets will be forwarded according to the bridging pipeline. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 428 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 429 | In addition, a **copy of the ARP packet will be sent to the controller**. |
| 430 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 431 | - Controller will use the ARP packets for **learning purpose and update host |
| 432 | store** accordingly. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 433 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 434 | - Controller only **replies** an ARP request if the request is trying to |
| 435 | **resolve an interface address configured on the switch edge port**. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 436 | |
| 437 | Pipeline Walkthrough - ARP |
| 438 | ^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 439 | |
| 440 | It is similar to L2 broadcast. Except ARP packets will be matched by a special |
| 441 | ACL table entry and being copied to the controller. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 442 | |
| 443 | |
| 444 | L3 Unicast |
| 445 | ---------- |
| 446 | |
| 447 | .. image:: images/arch-l3u.png |
| 448 | :width: 800px |
| 449 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 450 | Fig. 7: L3 unicast |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 451 | |
| 452 | .. image:: images/arch-l3u-src-pipeline.png |
| 453 | :width: 1000px |
| 454 | |
| 455 | Fig. 8 Simplified L3 unicast pipeline - source leaf |
| 456 | |
| 457 | Pipeline Walkthrough - Source Leaf Switch |
| 458 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 459 | |
| 460 | - **VLAN Table**: An untagged packet will be assigned an internal VLAN ID |
| 461 | according to the input port and the subnet configured on the input port. |
| 462 | Packets of the same subnet will have the same internal VLAN ID. |
| 463 | |
| 464 | - **TMAC Table**: Since the destination MAC of a L3 unicast packet is the MAC |
| 465 | of leaf router and the ethernet type is IPv4, the packet will match the TMAC |
| 466 | table and go to the unicast routing table. |
| 467 | |
| 468 | - **Unicast Routing Table**: In this table we will lookup the destination IP |
| 469 | of the packet and point the packet to corresponding L3 ECMP group |
| 470 | |
| 471 | - **ACL Table**: IP packets will miss the ACL table and the L3 ECMP group will |
| 472 | be executed. |
| 473 | |
| 474 | - **L3 ECMP Group**: Hashes on 5 tuple to pick a spine switch and goto the |
| 475 | MPLS Label Group. |
| 476 | |
| 477 | - **MPLS Label Group**: Push the MPLS label corresponding to the destination |
| 478 | leaf switch and goto the MPLS Interface Group. |
| 479 | |
| 480 | - **MPLS Interface Group**: Set source MAC address, destination MAC address, |
| 481 | VLAN ID and goto the L2 Interface Group. |
| 482 | |
| 483 | - **L2 Interface Group**: The internal assigned VLAN will be popped before the |
| 484 | packet is sent to the output port that goes to the spine. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 485 | |
| 486 | .. image:: images/arch-l3u-transit-pipeline.png |
| 487 | :width: 1000px |
| 488 | |
| 489 | Fig. 9 Simplified L3 unicast pipeline - spine |
| 490 | |
| 491 | Pipeline Walkthrough - Spine Switch |
| 492 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 493 | |
| 494 | - **VLAN Table**: An untagged packet will be assigned an internal VLAN ID |
| 495 | according to the input port and the subnet configured on the input port. |
| 496 | Packets of the same subnet will have the same internal VLAN ID. |
| 497 | |
| 498 | - **TMAC Table**: Since the destination MAC of a L3 unicast packet is the MAC |
| 499 | of spine router and the ethernet type is MPLS, the packet will match the TMAC |
| 500 | table and go to the MPLS table. |
| 501 | |
| 502 | - **MPLS Table**: In this table we will lookup the MPLS label of the packet, |
| 503 | figure out the destination leaf switch, pop the MPLS label and point to L3 |
| 504 | ECMP Group. |
| 505 | |
| 506 | - **ACL Table**: IP packets will miss the ACL table and the MPLS interface |
| 507 | group will be executed. |
| 508 | |
| 509 | - **L3 ECMP Group**: Hash to pick a link (if there are multiple links) to the |
| 510 | destination leaf and goto the L3 Interface Group. |
| 511 | |
| 512 | - **MPLS Interface Group**: Set source MAC address, destination MAC address, |
| 513 | VLAN ID and goto the L2 Interface Group. |
| 514 | |
| 515 | - **L2 Interface Group**: The internal assigned VLAN will be popped before the |
| 516 | packet is sent to the output port that goes to the destination leaf switch. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 517 | |
| 518 | .. image:: images/arch-l3u-dst-pipeline.png |
| 519 | :width: 1000px |
| 520 | |
| 521 | Fig. 10 Simplified L3 unicast pipeline - destination leaf |
| 522 | |
| 523 | Pipeline Walkthrough - Destination Leaf Switch |
| 524 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 525 | |
| 526 | - **VLAN Table**: An untagged packet will be assigned an internal VLAN ID |
| 527 | according to the input port and the subnet configured on the input port. |
| 528 | Packets of the same subnet will have the same internal VLAN ID. |
| 529 | |
| 530 | - **TMAC Table**: Since the destination MAC of a L3 unicast packet is the MAC |
| 531 | of leaf router and the ethernet type is IPv4, the packet will match the TMAC |
| 532 | table and go to the unicast routing table. |
| 533 | |
| 534 | - **Unicast Routing Table**: In this table we will lookup the destination IP |
| 535 | of the packet and point the packet to corresponding L3 Unicast Group. |
| 536 | |
| 537 | - **ACL Table**: IP packets will miss the ACL table and the L3 Unicast Group |
| 538 | will be executed. |
| 539 | |
| 540 | - **L3 Unicast Group**: Set source MAC address, destination MAC address, VLAN |
| 541 | ID and goto the L2 Interface Group. |
| 542 | |
| 543 | - **L2 Interface Group**: The internal assigned VLAN will be popped before the |
| 544 | packet is sent to the output port that goes to the destination leaf switch. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 545 | |
| 546 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 547 | The L3 unicast mechanism is designed to support inter-rack(inter-subnet) |
| 548 | untagged communication when the destination host is known. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 549 | |
| 550 | Path Calculation and Failover - Unicast |
| 551 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 552 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 553 | Coming soon... |
| 554 | |
| 555 | |
| 556 | L3 Multicast |
| 557 | ------------ |
| 558 | |
| 559 | .. image:: images/arch-l3m.png |
| 560 | :width: 800px |
| 561 | |
| 562 | Fig. 11 L3 multicast |
| 563 | |
| 564 | .. image:: images/arch-l3m-pipeline.png |
| 565 | :width: 1000px |
| 566 | |
| 567 | Fig.12 Simplified L3 multicast pipeline |
| 568 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 569 | The L3 multicast mechanism is designed to support use cases such as IPTV. The |
| 570 | multicast traffic comes in from the upstream router, replicated by the |
| 571 | leaf-spine switches, send to multiple OLTs and eventually get to the |
| 572 | subscribers. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 573 | |
| 574 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 575 | We would like to support different combinations of ingress/egress VLAN, |
| 576 | including |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 577 | |
| 578 | - untagged in -> untagged out |
| 579 | - untagged in -> tagged out |
| 580 | - tagged in -> untagged out |
| 581 | - tagged in -> same tagged out |
| 582 | - tagged in -> different tagged out |
| 583 | |
| 584 | However, due to the above-mentioned OFDPA restrictions, |
| 585 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 586 | - It is NOT possible to chain L3 multicast group to L2 interface group |
| 587 | directly if we want to change the VLAN ID |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 588 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 589 | - It is NOT possible to change VLAN ID by chaining L3 multicast group to L3 |
| 590 | interface group since all output ports should have the same VLAN but the |
| 591 | spec requires chained L3 interface group to have different VLAN ID from |
| 592 | each other. |
| 593 | |
| 594 | That means, if we need to change VLAN ID, we need to change it before the |
| 595 | packets get into the multicast routing table. |
| 596 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 597 | The only viable solution is changing the VLAN ID in the VLAN table. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 598 | |
| 599 | We change the VLAN tag on the ingress switch (i.e. the switch that connects |
| 600 | to the upstream router) when necessary. |
| 601 | |
| 602 | On transit (spine) and egress (destination leaf) switches, output VLAN tag |
| 603 | will remain the same as input VLAN tag. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 604 | |
| 605 | Pipeline Walkthrough - Ingress Leaf Switch |
| 606 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
| 607 | |
| 608 | .. csv-table:: Table 1. All Possible VLAN Combinations on Ingress Switch |
| 609 | :file: tables/arch-mcast-ingress.csv |
| 610 | :widths: 2, 5, 5, 10, 10, 5 |
| 611 | :header-rows: 1 |
| 612 | |
| 613 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 614 | In the presence of ``vlan-untagged`` configuration on the ingress port of |
| 615 | the ingress switch, the ``vlan-untagged`` will be used instead of 4094. |
| 616 | |
| 617 | The reason is that we cannot distinguish unicast and multicast traffic in |
| 618 | that case, and therefore must assign the same VLAN to the packet. |
| 619 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 620 | The VLAN will anyway get popped in L2IG in this case. |
| 621 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 622 | Table 1 shows all possible VLAN combinations on the ingress switches and how |
| 623 | the packet is processed through the pipeline. We take the second case |
| 624 | **untagged -> tagged 200** as an example to explain more details. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 625 | |
| 626 | - **VLAN Table**: An untagged packet will be assigned the **egress VLAN ID**. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 627 | |
| 628 | - **TMAC Table**: Since the destination MAC of a L2 unicast packet is a |
| 629 | multicast MAC address, the packet will match the TMAC table and goes to the |
| 630 | multicast routing table. |
| 631 | |
| 632 | - **Multicast Routing Table**: In this table we will lookup the multicast group |
| 633 | (destination multicast IP) and point the packet to the corresponding L3 |
| 634 | multicast group. |
| 635 | |
| 636 | - **ACL Table**: Multicast packets will miss the ACL table and the L3 multicast |
| 637 | group will be executed. |
| 638 | |
| 639 | - **L3 Multicast Group**: The packet will be matched by **egress VLAN ID** and |
| 640 | forwarded to multiple L2 interface groups that map to output ports. |
| 641 | |
| 642 | - **L2 Interface Group**: The egress VLAN will be kept in this case and the |
| 643 | packet will be sent to the output port that goes to the transit spine switch. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 644 | |
| 645 | |
| 646 | Pipeline Walkthrough - Transit Spine Switch and Egress Leaf Switch |
| 647 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
| 648 | |
| 649 | .. csv-table:: Table 2. All Possible VLAN Combinations on Transit/Egress Switch |
| 650 | :file: tables/arch-mcast-transit-egress.csv |
| 651 | :widths: 2, 5, 5, 10, 10, 5 |
| 652 | :header-rows: 1 |
| 653 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 654 | Table 2 shows all possible VLAN combinations on the transit/egress switches and |
| 655 | how the packet is processed through the pipeline. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 656 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 657 | Note that we have already changed the VLAN tag to the desired egress VLAN on |
| 658 | the ingress switch. |
| 659 | |
| 660 | Therefore, there are only two cases on the transit/egress switches - either |
| 661 | keep it untagged or keep it tagged. We take the first case **untagged -> |
| 662 | untagged** as an example to explain more details. |
| 663 | |
| 664 | |
| 665 | - **VLAN Table**: An untagged packet will be assigned an **internal VLAN ID** |
| 666 | according to the input port and the subnet configured on the input port. |
| 667 | Packets of the same subnet will have the same internal VLAN ID. |
| 668 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 669 | - **TMAC Table**: (same as ingress switch) |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 670 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 671 | - **Multicast Routing Table**: (same as ingress switch) |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 672 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 673 | - **ACL Table**: (same as ingress switch) |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 674 | |
| 675 | - **L3 Multicast Group**: The packet will be matched by **internal VLAN ID** |
| 676 | and forwarded to multiple L2 interface groups that map to output ports. |
| 677 | |
| 678 | - **L2 Interface Group**: The egress VLAN will be popped in this case and the |
| 679 | packet will be sent to the output port that goes to the egress leaf switch. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 680 | |
| 681 | Path Calculation and Failover - Multicast |
| 682 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
| 683 | Coming soon... |
| 684 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 685 | VLAN Cross Connect |
| 686 | ------------------ |
| 687 | |
| 688 | .. image:: images/arch-xconnect.png |
| 689 | :width: 800px |
| 690 | |
| 691 | Fig. 13 VLAN cross connect |
| 692 | |
| 693 | .. image:: images/arch-xconnect-pipeline.png |
| 694 | :width: 1000px |
| 695 | |
| 696 | Fig. 14 Simplified VLAN cross connect pipeline |
| 697 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 698 | VLAN Cross Connect is originally designed to support Q-in-Q packets between |
| 699 | OLTs and BNGs. |
| 700 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 701 | The cross connect pair consists of two output ports. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 702 | |
| 703 | Whatever packet comes in on one port with specific VLAN tag will be sent to the |
| 704 | other port. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 705 | |
| 706 | .. note:: |
| 707 | It can only cross connects **two ports on the same switch**. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 708 | :doc:`Pseudowire <configuration/pseudowire>` is required to connect ports |
| 709 | across different switches. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 710 | |
| 711 | We use L2 Flood Group to implement VLAN Cross Connect. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 712 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 713 | The L2 Flood Group for cross connect only consists of two ports. |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 714 | |
| 715 | The input port will be removed before flooding according to the spec and thus |
| 716 | create exactly the desire behavior of cross connect. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 717 | |
| 718 | Pipeline Walkthrough - Cross Connect |
| 719 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 720 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 721 | - **VLAN Table**: When a tagged packet comes in, we no longer need to assign |
| 722 | the internal VLAN. The original VLAN will be carried through the entire |
| 723 | pipeline. |
| 724 | |
| 725 | - **TMAC Table**: Since the VLAN will not match any internal VLAN assigned to |
| 726 | untagged packets, the packet will miss the TMAC table and goes to the |
| 727 | bridging table. |
| 728 | |
| 729 | - **Bridging Table**: The packet will hit the flow rule that match the cross |
| 730 | connect VLAN ID and being sent to corresponding L2 Flood Group. |
| 731 | |
| 732 | - **ACL Table**: IP packets will miss the ACL table and the L2 flood group will |
| 733 | be executed. |
| 734 | |
| 735 | - **L2 Flood Group**: Consists of two L2 interface groups related to this cross |
| 736 | connect VLAN. L2 Interface Group: The original VLAN will NOT be popped |
| 737 | before the packet is sent to the output port. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 738 | |
| 739 | vRouter |
| 740 | ------- |
| 741 | |
| 742 | .. image:: images/arch-vr.png |
| 743 | :width: 800px |
| 744 | |
| 745 | Fig. 15 vRouter |
| 746 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 747 | The Trellis fabric needs to be connected to the external world via the vRouter |
| 748 | functionality. **In the networking industry, the term vRouter implies a |
| 749 | "router in a VM". This is not the case in Trellis**. Trellis vRouter is NOT a |
| 750 | software router. |
| 751 | |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 752 | **Only the control plane of the router, i.e routing protocols, runs in a VM**. |
| 753 | We use the Quagga routing protocol suite as the control plane for vRouter. |
| 754 | |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 755 | The **vRouter data plane is entirely in hardware**. Essentially the entire |
| 756 | hardware fabric serves as the (distributed) data plane for vRouter. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 757 | |
| 758 | The **external router views the entire Trellis fabric as a single router**. |
| 759 | |
| 760 | .. image:: images/arch-vr-overview.png |
| 761 | |
| 762 | .. image:: images/arch-vr-logical.png |
| 763 | |
| 764 | .. note:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 765 | Dual external routers is also supported for redundancy. Visit |
| 766 | :doc:`External Connectivity <configuration/dual-homing>` for details. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 767 | |
| 768 | Pipeline Walkthrough - vRouter |
| 769 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 770 | |
| 771 | The pipeline is exactly as same as L3 unicast. We just install additional flow |
| 772 | rules in the unicast routing table on each leaf routers. |
Charles Chan | 33bac08 | 2019-09-12 01:07:51 -0700 | [diff] [blame] | 773 | |
| 774 | |
| 775 | Learn More |
| 776 | ---------- |
| 777 | .. tip:: |
Zack Williams | d63d35b | 2020-06-23 14:12:46 -0700 | [diff] [blame] | 778 | Most of our design discussion and meeting notes are kept in `Google Drive |
| 779 | <https://drive.google.com/drive/folders/0Bz9dNKPVvtgsR0M5R0hWSHlfZ0U>`_. |
| 780 | If you are wondering why features are designed and implemented in a certain |
| 781 | way, you may find the answers there. |