1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright(c) 2010-2014 Intel Corporation. 3 4Packet Framework 5================ 6 7Design Objectives 8----------------- 9 10The main design objectives for the DPDK Packet Framework are: 11 12* Provide standard methodology to build complex packet processing pipelines. 13 Provide reusable and extensible templates for the commonly used pipeline functional blocks; 14 15* Provide capability to switch between pure software and hardware-accelerated implementations for the same pipeline functional block; 16 17* Provide the best trade-off between flexibility and performance. 18 Hardcoded pipelines usually provide the best performance, but are not flexible, 19 while developing flexible frameworks is never a problem, but performance is usually low; 20 21* Provide a framework that is logically similar to Open Flow. 22 23Overview 24-------- 25 26Packet processing applications are frequently structured as pipelines of multiple stages, 27with the logic of each stage glued around a lookup table. 28For each incoming packet, the table defines the set of actions to be applied to the packet, 29as well as the next stage to send the packet to. 30 31The DPDK Packet Framework minimizes the development effort required to build packet processing pipelines 32by defining a standard methodology for pipeline development, 33as well as providing libraries of reusable templates for the commonly used pipeline blocks. 34 35The pipeline is constructed by connecting the set of input ports with the set of output ports 36through the set of tables in a tree-like topology. 37As result of lookup operation for the current packet in the current table, 38one of the table entries (on lookup hit) or the default table entry (on lookup miss) 39provides the set of actions to be applied on the current packet, 40as well as the next hop for the packet, which can be either another table, an output port or packet drop. 41 42An example of packet processing pipeline is presented in :numref:`figure_figure32`: 43 44.. _figure_figure32: 45 46.. figure:: img/figure32.* 47 48 Example of Packet Processing Pipeline where Input Ports 0 and 1 49 are Connected with Output Ports 0, 1 and 2 through Tables 0 and 1 50 51 52Port Library Design 53------------------- 54 55Port Types 56~~~~~~~~~~ 57 58:numref:`table_qos_19` is a non-exhaustive list of ports that can be implemented with the Packet Framework. 59 60.. _table_qos_19: 61 62.. table:: Port Types 63 64 +---+------------------+---------------------------------------------------------------------------------------+ 65 | # | Port type | Description | 66 | | | | 67 +===+==================+=======================================================================================+ 68 | 1 | SW ring | SW circular buffer used for message passing between the application threads. Uses | 69 | | | the DPDK rte_ring primitive. Expected to be the most commonly used type of | 70 | | | port. | 71 | | | | 72 +---+------------------+---------------------------------------------------------------------------------------+ 73 | 2 | HW ring | Queue of buffer descriptors used to interact with NIC, switch or accelerator ports. | 74 | | | For NIC ports, it uses the DPDK rte_eth_rx_queue or rte_eth_tx_queue | 75 | | | primitives. | 76 | | | | 77 +---+------------------+---------------------------------------------------------------------------------------+ 78 | 3 | IP reassembly | Input packets are either IP fragments or complete IP datagrams. Output packets are | 79 | | | complete IP datagrams. | 80 | | | | 81 +---+------------------+---------------------------------------------------------------------------------------+ 82 | 4 | IP fragmentation | Input packets are jumbo (IP datagrams with length bigger than MTU) or non-jumbo | 83 | | | packets. Output packets are non-jumbo packets. | 84 | | | | 85 +---+------------------+---------------------------------------------------------------------------------------+ 86 | 5 | Traffic manager | Traffic manager attached to a specific NIC output port, performing congestion | 87 | | | management and hierarchical scheduling according to pre-defined SLAs. | 88 | | | | 89 +---+------------------+---------------------------------------------------------------------------------------+ 90 | 6 | KNI | Send/receive packets to/from Linux kernel space. | 91 | | | | 92 +---+------------------+---------------------------------------------------------------------------------------+ 93 | 7 | Source | Input port used as packet generator. Similar to Linux kernel /dev/zero character | 94 | | | device. | 95 | | | | 96 +---+------------------+---------------------------------------------------------------------------------------+ 97 | 8 | Sink | Output port used to drop all input packets. Similar to Linux kernel /dev/null | 98 | | | character device. | 99 | | | | 100 +---+------------------+---------------------------------------------------------------------------------------+ 101 102Port Interface 103~~~~~~~~~~~~~~ 104 105Each port is unidirectional, i.e. either input port or output port. 106Each input/output port is required to implement an abstract interface that 107defines the initialization and run-time operation of the port. 108The port abstract interface is described in. 109 110.. _table_qos_20: 111 112.. table:: 20 Port Abstract Interface 113 114 +---+----------------+-----------------------------------------------------------------------------------------+ 115 | # | Port Operation | Description | 116 | | | | 117 +===+================+=========================================================================================+ 118 | 1 | Create | Create the low-level port object (e.g. queue). Can internally allocate memory. | 119 | | | | 120 +---+----------------+-----------------------------------------------------------------------------------------+ 121 | 2 | Free | Free the resources (e.g. memory) used by the low-level port object. | 122 | | | | 123 +---+----------------+-----------------------------------------------------------------------------------------+ 124 | 3 | RX | Read a burst of input packets. Non-blocking operation. Only defined for input ports. | 125 | | | | 126 +---+----------------+-----------------------------------------------------------------------------------------+ 127 | 4 | TX | Write a burst of input packets. Non-blocking operation. Only defined for output ports. | 128 | | | | 129 +---+----------------+-----------------------------------------------------------------------------------------+ 130 | 5 | Flush | Flush the output buffer. Only defined for output ports. | 131 | | | | 132 +---+----------------+-----------------------------------------------------------------------------------------+ 133 134Table Library Design 135-------------------- 136 137Table Types 138~~~~~~~~~~~ 139 140:numref:`table_qos_21` is a non-exhaustive list of types of tables that can be implemented with the Packet Framework. 141 142.. _table_qos_21: 143 144.. table:: Table Types 145 146 +---+----------------------------+-----------------------------------------------------------------------------+ 147 | # | Table Type | Description | 148 | | | | 149 +===+============================+=============================================================================+ 150 | 1 | Hash table | Lookup key is n-tuple based. | 151 | | | | 152 | | | Typically, the lookup key is hashed to produce a signature that is used to | 153 | | | identify a bucket of entries where the lookup key is searched next. | 154 | | | | 155 | | | The signature associated with the lookup key of each input packet is either | 156 | | | read from the packet descriptor (pre-computed signature) or computed at | 157 | | | table lookup time. | 158 | | | | 159 | | | The table lookup, add entry and delete entry operations, as well as any | 160 | | | other pipeline block that pre-computes the signature all have to use the | 161 | | | same hashing algorithm to generate the signature. | 162 | | | | 163 | | | Typically used to implement flow classification tables, ARP caches, routing | 164 | | | table for tunnelling protocols, etc. | 165 | | | | 166 +---+----------------------------+-----------------------------------------------------------------------------+ 167 | 2 | Longest Prefix Match (LPM) | Lookup key is the IP address. | 168 | | | | 169 | | | Each table entries has an associated IP prefix (IP and depth). | 170 | | | | 171 | | | The table lookup operation selects the IP prefix that is matched by the | 172 | | | lookup key; in case of multiple matches, the entry with the longest prefix | 173 | | | depth wins. | 174 | | | | 175 | | | Typically used to implement IP routing tables. | 176 | | | | 177 +---+----------------------------+-----------------------------------------------------------------------------+ 178 | 3 | Access Control List (ACLs) | Lookup key is 7-tuple of two VLAN/MPLS labels, IP destination address, | 179 | | | IP source addresses, L4 protocol, L4 destination port, L4 source port. | 180 | | | | 181 | | | Each table entry has an associated ACL and priority. The ACL contains bit | 182 | | | masks for the VLAN/MPLS labels, IP prefix for IP destination address, IP | 183 | | | prefix for IP source addresses, L4 protocol and bitmask, L4 destination | 184 | | | port and bit mask, L4 source port and bit mask. | 185 | | | | 186 | | | The table lookup operation selects the ACL that is matched by the lookup | 187 | | | key; in case of multiple matches, the entry with the highest priority wins. | 188 | | | | 189 | | | Typically used to implement rule databases for firewalls, etc. | 190 | | | | 191 +---+----------------------------+-----------------------------------------------------------------------------+ 192 | 4 | Pattern matching search | Lookup key is the packet payload. | 193 | | | | 194 | | | Table is a database of patterns, with each pattern having a priority | 195 | | | assigned. | 196 | | | | 197 | | | The table lookup operation selects the patterns that is matched by the | 198 | | | input packet; in case of multiple matches, the matching pattern with the | 199 | | | highest priority wins. | 200 | | | | 201 +---+----------------------------+-----------------------------------------------------------------------------+ 202 | 5 | Array | Lookup key is the table entry index itself. | 203 | | | | 204 +---+----------------------------+-----------------------------------------------------------------------------+ 205 206Table Interface 207~~~~~~~~~~~~~~~ 208 209Each table is required to implement an abstract interface that defines the initialization 210and run-time operation of the table. 211The table abstract interface is described in :numref:`table_qos_29_1`. 212 213.. _table_qos_29_1: 214 215.. table:: Table Abstract Interface 216 217 +---+-----------------+----------------------------------------------------------------------------------------+ 218 | # | Table operation | Description | 219 | | | | 220 +===+=================+========================================================================================+ 221 | 1 | Create | Create the low-level data structures of the lookup table. Can internally allocate | 222 | | | memory. | 223 | | | | 224 +---+-----------------+----------------------------------------------------------------------------------------+ 225 | 2 | Free | Free up all the resources used by the lookup table. | 226 | | | | 227 +---+-----------------+----------------------------------------------------------------------------------------+ 228 | 3 | Add entry | Add new entry to the lookup table. | 229 | | | | 230 +---+-----------------+----------------------------------------------------------------------------------------+ 231 | 4 | Delete entry | Delete specific entry from the lookup table. | 232 | | | | 233 +---+-----------------+----------------------------------------------------------------------------------------+ 234 | 5 | Lookup | Look up a burst of input packets and return a bit mask specifying the result of the | 235 | | | lookup operation for each packet: a set bit signifies lookup hit for the corresponding | 236 | | | packet, while a cleared bit a lookup miss. | 237 | | | | 238 | | | For each lookup hit packet, the lookup operation also returns a pointer to the table | 239 | | | entry that was hit, which contains the actions to be applied on the packet and any | 240 | | | associated metadata. | 241 | | | | 242 | | | For each lookup miss packet, the actions to be applied on the packet and any | 243 | | | associated metadata are specified by the default table entry preconfigured for lookup | 244 | | | miss. | 245 | | | | 246 +---+-----------------+----------------------------------------------------------------------------------------+ 247 248 249Hash Table Design 250~~~~~~~~~~~~~~~~~ 251 252Hash Table Overview 253^^^^^^^^^^^^^^^^^^^ 254 255Hash tables are important because the key lookup operation is optimized for speed: 256instead of having to linearly search the lookup key through all the keys in the table, 257the search is limited to only the keys stored in a single table bucket. 258 259**Associative Arrays** 260 261An associative array is a function that can be specified as a set of (key, value) pairs, 262with each key from the possible set of input keys present at most once. 263For a given associative array, the possible operations are: 264 265#. *add (key, value)*: When no value is currently associated with *key*, then the (key, *value* ) association is created. 266 When *key* is already associated value *value0*, then the association (*key*, *value0*) is removed 267 and association *(key, value)* is created; 268 269#. *delete key*: When no value is currently associated with *key*, this operation has no effect. 270 When *key* is already associated *value*, then association *(key, value)* is removed; 271 272#. *lookup key*: When no value is currently associated with *key*, then this operation returns void value (lookup miss). 273 When *key* is associated with *value*, then this operation returns *value*. 274 The *(key, value)* association is not changed. 275 276The matching criterion used to compare the input key against the keys in the associative array is *exact match*, 277as the key size (number of bytes) and the key value (array of bytes) have to match exactly for the two keys under comparison. 278 279**Hash Function** 280 281A hash function deterministically maps data of variable length (key) to data of fixed size (hash value or key signature). 282Typically, the size of the key is bigger than the size of the key signature. 283The hash function basically compresses a long key into a short signature. 284Several keys can share the same signature (collisions). 285 286High quality hash functions have uniform distribution. 287For large number of keys, when dividing the space of signature values into a fixed number of equal intervals (buckets), 288it is desirable to have the key signatures evenly distributed across these intervals (uniform distribution), 289as opposed to most of the signatures going into only a few of the intervals 290and the rest of the intervals being largely unused (non-uniform distribution). 291 292**Hash Table** 293 294A hash table is an associative array that uses a hash function for its operation. 295The reason for using a hash function is to optimize the performance of the lookup operation 296by minimizing the number of table keys that have to be compared against the input key. 297 298Instead of storing the (key, value) pairs in a single list, the hash table maintains multiple lists (buckets). 299For any given key, there is a single bucket where that key might exist, and this bucket is uniquely identified based on the key signature. 300Once the key signature is computed and the hash table bucket identified, 301the key is either located in this bucket or it is not present in the hash table at all, 302so the key search can be narrowed down from the full set of keys currently in the table 303to just the set of keys currently in the identified table bucket. 304 305The performance of the hash table lookup operation is greatly improved, 306provided that the table keys are evenly distributed among the hash table buckets, 307which can be achieved by using a hash function with uniform distribution. 308The rule to map a key to its bucket can simply be to use the key signature (modulo the number of table buckets) as the table bucket ID: 309 310 *bucket_id = f_hash(key) % n_buckets;* 311 312By selecting the number of buckets to be a power of two, the modulo operator can be replaced by a bitwise AND logical operation: 313 314 *bucket_id = f_hash(key) & (n_buckets - 1);* 315 316considering *n_bits* as the number of bits set in *bucket_mask = n_buckets - 1*, 317this means that all the keys that end up in the same hash table bucket have the lower *n_bits* of their signature identical. 318In order to reduce the number of keys in the same bucket (collisions), the number of hash table buckets needs to be increased. 319 320In packet processing context, the sequence of operations involved in hash table operations is described in :numref:`figure_figure33`: 321 322.. _figure_figure33: 323 324.. figure:: img/figure33.* 325 326 Sequence of Steps for Hash Table Operations in a Packet Processing Context 327 328 329 330Hash Table Use Cases 331^^^^^^^^^^^^^^^^^^^^ 332 333**Flow Classification** 334 335*Description:* The flow classification is executed at least once for each input packet. 336This operation maps each incoming packet against one of the known traffic flows in the flow database that typically contains millions of flows. 337 338*Hash table name:* Flow classification table 339 340*Number of keys:* Millions 341 342*Key format:* n-tuple of packet fields that uniquely identify a traffic flow/connection. 343Example: DiffServ 5-tuple of (Source IP address, Destination IP address, L4 protocol, L4 protocol source port, L4 protocol destination port). 344For IPv4 protocol and L4 protocols like TCP, UDP or SCTP, the size of the DiffServ 5-tuple is 13 bytes, while for IPv6 it is 37 bytes. 345 346*Key value (key data):* actions and action meta-data describing what processing to be applied for the packets of the current flow. 347The size of the data associated with each traffic flow can vary from 8 bytes to kilobytes. 348 349**Address Resolution Protocol (ARP)** 350 351*Description:* Once a route has been identified for an IP packet (so the output interface and the IP address of the next hop station are known), 352the MAC address of the next hop station is needed in order to send this packet onto the next leg of the journey 353towards its destination (as identified by its destination IP address). 354The MAC address of the next hop station becomes the destination MAC address of the outgoing Ethernet frame. 355 356*Hash table name:* ARP table 357 358*Number of keys:* Thousands 359 360*Key format:* The pair of (Output interface, Next Hop IP address), which is typically 5 bytes for IPv4 and 17 bytes for IPv6. 361 362*Key value (key data):* MAC address of the next hop station (6 bytes). 363 364Hash Table Types 365^^^^^^^^^^^^^^^^ 366 367:numref:`table_qos_22` lists the hash table configuration parameters shared by all different hash table types. 368 369.. _table_qos_22: 370 371.. table:: Configuration Parameters Common for All Hash Table Types 372 373 +---+---------------------------+------------------------------------------------------------------------------+ 374 | # | Parameter | Details | 375 | | | | 376 +===+===========================+==============================================================================+ 377 | 1 | Key size | Measured as number of bytes. All keys have the same size. | 378 | | | | 379 +---+---------------------------+------------------------------------------------------------------------------+ 380 | 2 | Key value (key data) size | Measured as number of bytes. | 381 | | | | 382 +---+---------------------------+------------------------------------------------------------------------------+ 383 | 3 | Number of buckets | Needs to be a power of two. | 384 | | | | 385 +---+---------------------------+------------------------------------------------------------------------------+ 386 | 4 | Maximum number of keys | Needs to be a power of two. | 387 | | | | 388 +---+---------------------------+------------------------------------------------------------------------------+ 389 | 5 | Hash function | Examples: jhash, CRC hash, etc. | 390 | | | | 391 +---+---------------------------+------------------------------------------------------------------------------+ 392 | 6 | Hash function seed | Parameter to be passed to the hash function. | 393 | | | | 394 +---+---------------------------+------------------------------------------------------------------------------+ 395 | 7 | Key offset | Offset of the lookup key byte array within the packet meta-data stored in | 396 | | | the packet buffer. | 397 | | | | 398 +---+---------------------------+------------------------------------------------------------------------------+ 399 400Bucket Full Problem 401""""""""""""""""""" 402 403On initialization, each hash table bucket is allocated space for exactly 4 keys. 404As keys are added to the table, it can happen that a given bucket already has 4 keys when a new key has to be added to this bucket. 405The possible options are: 406 407#. **Least Recently Used (LRU) Hash Table.** 408 One of the existing keys in the bucket is deleted and the new key is added in its place. 409 The number of keys in each bucket never grows bigger than 4. The logic to pick the key to be dropped from the bucket is LRU. 410 The hash table lookup operation maintains the order in which the keys in the same bucket are hit, so every time a key is hit, 411 it becomes the new Most Recently Used (MRU) key, i.e. the last candidate for drop. 412 When a key is added to the bucket, it also becomes the new MRU key. 413 When a key needs to be picked and dropped, the first candidate for drop, i.e. the current LRU key, is always picked. 414 The LRU logic requires maintaining specific data structures per each bucket. 415 416#. **Extendable Bucket Hash Table.** 417 The bucket is extended with space for 4 more keys. 418 This is done by allocating additional memory at table initialization time, 419 which is used to create a pool of free keys (the size of this pool is configurable and always a multiple of 4). 420 On key add operation, the allocation of a group of 4 keys only happens successfully within the limit of free keys, 421 otherwise the key add operation fails. 422 On key delete operation, a group of 4 keys is freed back to the pool of free keys 423 when the key to be deleted is the only key that was used within its group of 4 keys at that time. 424 On key lookup operation, if the current bucket is in extended state and a match is not found in the first group of 4 keys, 425 the search continues beyond the first group of 4 keys, potentially until all keys in this bucket are examined. 426 The extendable bucket logic requires maintaining specific data structures per table and per each bucket. 427 428.. _table_qos_23: 429 430.. table:: Configuration Parameters Specific to Extendable Bucket Hash Table 431 432 +---+---------------------------+--------------------------------------------------+ 433 | # | Parameter | Details | 434 | | | | 435 +===+===========================+==================================================+ 436 | 1 | Number of additional keys | Needs to be a power of two, at least equal to 4. | 437 | | | | 438 +---+---------------------------+--------------------------------------------------+ 439 440 441Signature Computation 442""""""""""""""""""""" 443 444The possible options for key signature computation are: 445 446#. **Pre-computed key signature.** 447 The key lookup operation is split between two CPU cores. 448 The first CPU core (typically the CPU core that performs packet RX) extracts the key from the input packet, 449 computes the key signature and saves both the key and the key signature in the packet buffer as packet meta-data. 450 The second CPU core reads both the key and the key signature from the packet meta-data 451 and performs the bucket search step of the key lookup operation. 452 453#. **Key signature computed on lookup ("do-sig" version).** 454 The same CPU core reads the key from the packet meta-data, uses it to compute the key signature 455 and also performs the bucket search step of the key lookup operation. 456 457.. _table_qos_24: 458 459.. table:: Configuration Parameters Specific to Pre-computed Key Signature Hash Table 460 461 +---+------------------+-----------------------------------------------------------------------+ 462 | # | Parameter | Details | 463 | | | | 464 +===+==================+=======================================================================+ 465 | 1 | Signature offset | Offset of the pre-computed key signature within the packet meta-data. | 466 | | | | 467 +---+------------------+-----------------------------------------------------------------------+ 468 469Key Size Optimized Hash Tables 470"""""""""""""""""""""""""""""" 471 472For specific key sizes, the data structures and algorithm of key lookup operation can be specially handcrafted for further performance improvements, 473so following options are possible: 474 475#. **Implementation supporting configurable key size.** 476 477#. **Implementation supporting a single key size.** 478 Typical key sizes are 8 bytes and 16 bytes. 479 480Bucket Search Logic for Configurable Key Size Hash Tables 481^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 482 483The performance of the bucket search logic is one of the main factors influencing the performance of the key lookup operation. 484The data structures and algorithm are designed to make the best use of Intel CPU architecture resources like: 485cache memory space, cache memory bandwidth, external memory bandwidth, multiple execution units working in parallel, 486out of order instruction execution, special CPU instructions, etc. 487 488The bucket search logic handles multiple input packets in parallel. 489It is built as a pipeline of several stages (3 or 4), with each pipeline stage handling two different packets from the burst of input packets. 490On each pipeline iteration, the packets are pushed to the next pipeline stage: for the 4-stage pipeline, 491two packets (that just completed stage 3) exit the pipeline, 492two packets (that just completed stage 2) are now executing stage 3, two packets (that just completed stage 1) are now executing stage 2, 493two packets (that just completed stage 0) are now executing stage 1 and two packets (next two packets to read from the burst of input packets) 494are entering the pipeline to execute stage 0. 495The pipeline iterations continue until all packets from the burst of input packets execute the last stage of the pipeline. 496 497The bucket search logic is broken into pipeline stages at the boundary of the next memory access. 498Each pipeline stage uses data structures that are stored (with high probability) into the L1 or L2 cache memory of the current CPU core and 499breaks just before the next memory access required by the algorithm. 500The current pipeline stage finalizes by prefetching the data structures required by the next pipeline stage, 501so given enough time for the prefetch to complete, 502when the next pipeline stage eventually gets executed for the same packets, 503it will read the data structures it needs from L1 or L2 cache memory and thus avoid the significant penalty incurred by L2 or L3 cache memory miss. 504 505By prefetching the data structures required by the next pipeline stage in advance (before they are used) 506and switching to executing another pipeline stage for different packets, 507the number of L2 or L3 cache memory misses is greatly reduced, hence one of the main reasons for improved performance. 508This is because the cost of L2/L3 cache memory miss on memory read accesses is high, as usually due to data dependency between instructions, 509the CPU execution units have to stall until the read operation is completed from L3 cache memory or external DRAM memory. 510By using prefetch instructions, the latency of memory read accesses is hidden, 511provided that it is preformed early enough before the respective data structure is actually used. 512 513By splitting the processing into several stages that are executed on different packets (the packets from the input burst are interlaced), 514enough work is created to allow the prefetch instructions to complete successfully (before the prefetched data structures are actually accessed) and 515also the data dependency between instructions is loosened. 516For example, for the 4-stage pipeline, stage 0 is executed on packets 0 and 1 and then, 517before same packets 0 and 1 are used (i.e. before stage 1 is executed on packets 0 and 1), 518different packets are used: packets 2 and 3 (executing stage 1), packets 4 and 5 (executing stage 2) and packets 6 and 7 (executing stage 3). 519By executing useful work while the data structures are brought into the L1 or L2 cache memory, the latency of the read memory accesses is hidden. 520By increasing the gap between two consecutive accesses to the same data structure, the data dependency between instructions is loosened; 521this allows making the best use of the super-scalar and out-of-order execution CPU architecture, 522as the number of CPU core execution units that are active (rather than idle or stalled due to data dependency constraints between instructions) is maximized. 523 524The bucket search logic is also implemented without using any branch instructions. 525This avoids the important cost associated with flushing the CPU core execution pipeline on every instance of branch misprediction. 526 527Configurable Key Size Hash Table 528"""""""""""""""""""""""""""""""" 529 530:numref:`figure_figure34`, :numref:`table_qos_25` and :numref:`table_qos_26` detail the main data structures used to implement configurable key size hash tables (either LRU or extendable bucket, 531either with pre-computed signature or "do-sig"). 532 533.. _figure_figure34: 534 535.. figure:: img/figure34.* 536 537 Data Structures for Configurable Key Size Hash Tables 538 539 540.. _table_qos_25: 541 542.. table:: Main Large Data Structures (Arrays) used for Configurable Key Size Hash Tables 543 544 +---+-------------------------+------------------------------+---------------------------+-------------------------------+ 545 | # | Array name | Number of entries | Entry size (bytes) | Description | 546 | | | | | | 547 +===+=========================+==============================+===========================+===============================+ 548 | 1 | Bucket array | n_buckets (configurable) | 32 | Buckets of the hash table. | 549 | | | | | | 550 +---+-------------------------+------------------------------+---------------------------+-------------------------------+ 551 | 2 | Bucket extensions array | n_buckets_ext (configurable) | 32 | This array is only created | 552 | | | | | for extendable bucket tables. | 553 | | | | | | 554 +---+-------------------------+------------------------------+---------------------------+-------------------------------+ 555 | 3 | Key array | n_keys | key_size (configurable) | Keys added to the hash table. | 556 | | | | | | 557 +---+-------------------------+------------------------------+---------------------------+-------------------------------+ 558 | 4 | Data array | n_keys | entry_size (configurable) | Key values (key data) | 559 | | | | | associated with the hash | 560 | | | | | table keys. | 561 | | | | | | 562 +---+-------------------------+------------------------------+---------------------------+-------------------------------+ 563 564.. _table_qos_26: 565 566.. table:: Field Description for Bucket Array Entry (Configurable Key Size Hash Tables) 567 568 +---+------------------+--------------------+------------------------------------------------------------------+ 569 | # | Field name | Field size (bytes) | Description | 570 | | | | | 571 +===+==================+====================+==================================================================+ 572 | 1 | Next Ptr/LRU | 8 | For LRU tables, this fields represents the LRU list for the | 573 | | | | current bucket stored as array of 4 entries of 2 bytes each. | 574 | | | | Entry 0 stores the index (0 .. 3) of the MRU key, while entry 3 | 575 | | | | stores the index of the LRU key. | 576 | | | | | 577 | | | | For extendable bucket tables, this field represents the next | 578 | | | | pointer (i.e. the pointer to the next group of 4 keys linked to | 579 | | | | the current bucket). The next pointer is not NULL if the bucket | 580 | | | | is currently extended or NULL otherwise. | 581 | | | | To help the branchless implementation, bit 0 (least significant | 582 | | | | bit) of this field is set to 1 if the next pointer is not NULL | 583 | | | | and to 0 otherwise. | 584 | | | | | 585 +---+------------------+--------------------+------------------------------------------------------------------+ 586 | 2 | Sig[0 .. 3] | 4 x 2 | If key X (X = 0 .. 3) is valid, then sig X bits 15 .. 1 store | 587 | | | | the most significant 15 bits of key X signature and sig X bit 0 | 588 | | | | is set to 1. | 589 | | | | | 590 | | | | If key X is not valid, then sig X is set to zero. | 591 | | | | | 592 +---+------------------+--------------------+------------------------------------------------------------------+ 593 | 3 | Key Pos [0 .. 3] | 4 x 4 | If key X is valid (X = 0 .. 3), then Key Pos X represents the | 594 | | | | index into the key array where key X is stored, as well as the | 595 | | | | index into the data array where the value associated with key X | 596 | | | | is stored. | 597 | | | | | 598 | | | | If key X is not valid, then the value of Key Pos X is undefined. | 599 | | | | | 600 +---+------------------+--------------------+------------------------------------------------------------------+ 601 602 603:numref:`figure_figure35` and :numref:`table_qos_27` detail the bucket search pipeline stages (either LRU or extendable bucket, 604either with pre-computed signature or "do-sig"). 605For each pipeline stage, the described operations are applied to each of the two packets handled by that stage. 606 607.. _figure_figure35: 608 609.. figure:: img/figure35.* 610 611 Bucket Search Pipeline for Key Lookup Operation (Configurable Key Size Hash 612 Tables) 613 614 615.. _table_qos_27: 616 617.. table:: Description of the Bucket Search Pipeline Stages (Configurable Key Size Hash Tables) 618 619 +---+---------------------------+------------------------------------------------------------------------------+ 620 | # | Stage name | Description | 621 | | | | 622 +===+===========================+==============================================================================+ 623 | 0 | Prefetch packet meta-data | Select next two packets from the burst of input packets. | 624 | | | | 625 | | | Prefetch packet meta-data containing the key and key signature. | 626 | | | | 627 +---+---------------------------+------------------------------------------------------------------------------+ 628 | 1 | Prefetch table bucket | Read the key signature from the packet meta-data (for extendable bucket hash | 629 | | | tables) or read the key from the packet meta-data and compute key signature | 630 | | | (for LRU tables). | 631 | | | | 632 | | | Identify the bucket ID using the key signature. | 633 | | | | 634 | | | Set bit 0 of the signature to 1 (to match only signatures of valid keys from | 635 | | | the table). | 636 | | | | 637 | | | Prefetch the bucket. | 638 | | | | 639 +---+---------------------------+------------------------------------------------------------------------------+ 640 | 2 | Prefetch table key | Read the key signatures from the bucket. | 641 | | | | 642 | | | Compare the signature of the input key against the 4 key signatures from the | 643 | | | packet. As result, the following is obtained: | 644 | | | | 645 | | | *match* | 646 | | | = equal to TRUE if there was at least one signature match and to FALSE in | 647 | | | the case of no signature match; | 648 | | | | 649 | | | *match_many* | 650 | | | = equal to TRUE is there were more than one signature matches (can be up to | 651 | | | 4 signature matches in the worst case scenario) and to FALSE otherwise; | 652 | | | | 653 | | | *match_pos* | 654 | | | = the index of the first key that produced signature match (only valid if | 655 | | | match is true). | 656 | | | | 657 | | | For extendable bucket hash tables only, set | 658 | | | *match_many* | 659 | | | to TRUE if next pointer is valid. | 660 | | | | 661 | | | Prefetch the bucket key indicated by | 662 | | | *match_pos* | 663 | | | (even if | 664 | | | *match_pos* | 665 | | | does not point to valid key valid). | 666 | | | | 667 +---+---------------------------+------------------------------------------------------------------------------+ 668 | 3 | Prefetch table data | Read the bucket key indicated by | 669 | | | *match_pos*. | 670 | | | | 671 | | | Compare the bucket key against the input key. As result, the following is | 672 | | | obtained: | 673 | | | *match_key* | 674 | | | = equal to TRUE if the two keys match and to FALSE otherwise. | 675 | | | | 676 | | | Report input key as lookup hit only when both | 677 | | | *match* | 678 | | | and | 679 | | | *match_key* | 680 | | | are equal to TRUE and as lookup miss otherwise. | 681 | | | | 682 | | | For LRU tables only, use branchless logic to update the bucket LRU list | 683 | | | (the current key becomes the new MRU) only on lookup hit. | 684 | | | | 685 | | | Prefetch the key value (key data) associated with the current key (to avoid | 686 | | | branches, this is done on both lookup hit and miss). | 687 | | | | 688 +---+---------------------------+------------------------------------------------------------------------------+ 689 690 691Additional notes: 692 693#. The pipelined version of the bucket search algorithm is executed only if there are at least 7 packets in the burst of input packets. 694 If there are less than 7 packets in the burst of input packets, 695 a non-optimized implementation of the bucket search algorithm is executed. 696 697#. Once the pipelined version of the bucket search algorithm has been executed for all the packets in the burst of input packets, 698 the non-optimized implementation of the bucket search algorithm is also executed for any packets that did not produce a lookup hit, 699 but have the *match_many* flag set. 700 As result of executing the non-optimized version, some of these packets may produce a lookup hit or lookup miss. 701 This does not impact the performance of the key lookup operation, 702 as the probability of matching more than one signature in the same group of 4 keys or of having the bucket in extended state 703 (for extendable bucket hash tables only) is relatively small. 704 705**Key Signature Comparison Logic** 706 707The key signature comparison logic is described in :numref:`table_qos_28`. 708 709.. _table_qos_28: 710 711.. table:: Lookup Tables for Match, Match_Many and Match_Pos 712 713 +----+------+---------------+--------------------+--------------------+ 714 | # | mask | match (1 bit) | match_many (1 bit) | match_pos (2 bits) | 715 | | | | | | 716 +----+------+---------------+--------------------+--------------------+ 717 | 0 | 0000 | 0 | 0 | 00 | 718 | | | | | | 719 +----+------+---------------+--------------------+--------------------+ 720 | 1 | 0001 | 1 | 0 | 00 | 721 | | | | | | 722 +----+------+---------------+--------------------+--------------------+ 723 | 2 | 0010 | 1 | 0 | 01 | 724 | | | | | | 725 +----+------+---------------+--------------------+--------------------+ 726 | 3 | 0011 | 1 | 1 | 00 | 727 | | | | | | 728 +----+------+---------------+--------------------+--------------------+ 729 | 4 | 0100 | 1 | 0 | 10 | 730 | | | | | | 731 +----+------+---------------+--------------------+--------------------+ 732 | 5 | 0101 | 1 | 1 | 00 | 733 | | | | | | 734 +----+------+---------------+--------------------+--------------------+ 735 | 6 | 0110 | 1 | 1 | 01 | 736 | | | | | | 737 +----+------+---------------+--------------------+--------------------+ 738 | 7 | 0111 | 1 | 1 | 00 | 739 | | | | | | 740 +----+------+---------------+--------------------+--------------------+ 741 | 8 | 1000 | 1 | 0 | 11 | 742 | | | | | | 743 +----+------+---------------+--------------------+--------------------+ 744 | 9 | 1001 | 1 | 1 | 00 | 745 | | | | | | 746 +----+------+---------------+--------------------+--------------------+ 747 | 10 | 1010 | 1 | 1 | 01 | 748 | | | | | | 749 +----+------+---------------+--------------------+--------------------+ 750 | 11 | 1011 | 1 | 1 | 00 | 751 | | | | | | 752 +----+------+---------------+--------------------+--------------------+ 753 | 12 | 1100 | 1 | 1 | 10 | 754 | | | | | | 755 +----+------+---------------+--------------------+--------------------+ 756 | 13 | 1101 | 1 | 1 | 00 | 757 | | | | | | 758 +----+------+---------------+--------------------+--------------------+ 759 | 14 | 1110 | 1 | 1 | 01 | 760 | | | | | | 761 +----+------+---------------+--------------------+--------------------+ 762 | 15 | 1111 | 1 | 1 | 00 | 763 | | | | | | 764 +----+------+---------------+--------------------+--------------------+ 765 766The input *mask* hash bit X (X = 0 .. 3) set to 1 if input signature is equal to bucket signature X and set to 0 otherwise. 767The outputs *match*, *match_many* and *match_pos* are 1 bit, 1 bit and 2 bits in size respectively and their meaning has been explained above. 768 769As displayed in :numref:`table_qos_29`, the lookup tables for *match* and *match_many* can be collapsed into a single 32-bit value and the lookup table for 770*match_pos* can be collapsed into a 64-bit value. 771Given the input *mask*, the values for *match*, *match_many* and *match_pos* can be obtained by indexing their respective bit array to extract 1 bit, 7721 bit and 2 bits respectively with branchless logic. 773 774.. _table_qos_29: 775 776.. table:: Collapsed Lookup Tables for Match, Match_Many and Match_Pos 777 778 +------------+------------------------------------------+-------------------+ 779 | | Bit array | Hexadecimal value | 780 | | | | 781 +------------+------------------------------------------+-------------------+ 782 | match | 1111_1111_1111_1110 | 0xFFFELLU | 783 | | | | 784 +------------+------------------------------------------+-------------------+ 785 | match_many | 1111_1110_1110_1000 | 0xFEE8LLU | 786 | | | | 787 +------------+------------------------------------------+-------------------+ 788 | match_pos | 0001_0010_0001_0011__0001_0010_0001_0000 | 0x12131210LLU | 789 | | | | 790 +------------+------------------------------------------+-------------------+ 791 792 793The pseudo-code for match, match_many and match_pos is:: 794 795 match = (0xFFFELLU >> mask) & 1; 796 797 match_many = (0xFEE8LLU >> mask) & 1; 798 799 match_pos = (0x12131210LLU >> (mask << 1)) & 3; 800 801Single Key Size Hash Tables 802""""""""""""""""""""""""""" 803 804:numref:`figure_figure37`, :numref:`figure_figure38`, :numref:`table_qos_30` and :numref:`table_qos_31` detail the main data structures used to implement 8-byte and 16-byte key hash tables 805(either LRU or extendable bucket, either with pre-computed signature or "do-sig"). 806 807.. _figure_figure37: 808 809.. figure:: img/figure37.* 810 811 Data Structures for 8-byte Key Hash Tables 812 813 814.. _figure_figure38: 815 816.. figure:: img/figure38.* 817 818 Data Structures for 16-byte Key Hash Tables 819 820 821.. _table_qos_30: 822 823.. table:: Main Large Data Structures (Arrays) used for 8-byte and 16-byte Key Size Hash Tables 824 825 +---+-------------------------+------------------------------+----------------------+------------------------------------+ 826 | # | Array name | Number of entries | Entry size (bytes) | Description | 827 | | | | | | 828 +===+=========================+==============================+======================+====================================+ 829 | 1 | Bucket array | n_buckets (configurable) | *8-byte key size:* | Buckets of the hash table. | 830 | | | | | | 831 | | | | 64 + 4 x entry_size | | 832 | | | | | | 833 | | | | | | 834 | | | | *16-byte key size:* | | 835 | | | | | | 836 | | | | 128 + 4 x entry_size | | 837 | | | | | | 838 +---+-------------------------+------------------------------+----------------------+------------------------------------+ 839 | 2 | Bucket extensions array | n_buckets_ext (configurable) | *8-byte key size:* | This array is only created for | 840 | | | | | extendable bucket tables. | 841 | | | | | | 842 | | | | 64 + 4 x entry_size | | 843 | | | | | | 844 | | | | | | 845 | | | | *16-byte key size:* | | 846 | | | | | | 847 | | | | 128 + 4 x entry_size | | 848 | | | | | | 849 +---+-------------------------+------------------------------+----------------------+------------------------------------+ 850 851.. _table_qos_31: 852 853.. table:: Field Description for Bucket Array Entry (8-byte and 16-byte Key Hash Tables) 854 855 +---+---------------+--------------------+-------------------------------------------------------------------------------+ 856 | # | Field name | Field size (bytes) | Description | 857 | | | | | 858 +===+===============+====================+===============================================================================+ 859 | 1 | Valid | 8 | Bit X (X = 0 .. 3) is set to 1 if key X is valid or to 0 otherwise. | 860 | | | | | 861 | | | | Bit 4 is only used for extendable bucket tables to help with the | 862 | | | | implementation of the branchless logic. In this case, bit 4 is set to 1 if | 863 | | | | next pointer is valid (not NULL) or to 0 otherwise. | 864 | | | | | 865 +---+---------------+--------------------+-------------------------------------------------------------------------------+ 866 | 2 | Next Ptr/LRU | 8 | For LRU tables, this fields represents the LRU list for the current bucket | 867 | | | | stored as array of 4 entries of 2 bytes each. Entry 0 stores the index | 868 | | | | (0 .. 3) of the MRU key, while entry 3 stores the index of the LRU key. | 869 | | | | | 870 | | | | For extendable bucket tables, this field represents the next pointer (i.e. | 871 | | | | the pointer to the next group of 4 keys linked to the current bucket). The | 872 | | | | next pointer is not NULL if the bucket is currently extended or NULL | 873 | | | | otherwise. | 874 | | | | | 875 +---+---------------+--------------------+-------------------------------------------------------------------------------+ 876 | 3 | Key [0 .. 3] | 4 x key_size | Full keys. | 877 | | | | | 878 +---+---------------+--------------------+-------------------------------------------------------------------------------+ 879 | 4 | Data [0 .. 3] | 4 x entry_size | Full key values (key data) associated with keys 0 .. 3. | 880 | | | | | 881 +---+---------------+--------------------+-------------------------------------------------------------------------------+ 882 883and detail the bucket search pipeline used to implement 8-byte and 16-byte key hash tables (either LRU or extendable bucket, 884either with pre-computed signature or "do-sig"). 885For each pipeline stage, the described operations are applied to each of the two packets handled by that stage. 886 887.. _figure_figure39: 888 889.. figure:: img/figure39.* 890 891 Bucket Search Pipeline for Key Lookup Operation (Single Key Size Hash 892 Tables) 893 894 895.. _table_qos_32: 896 897.. table:: Description of the Bucket Search Pipeline Stages (8-byte and 16-byte Key Hash Tables) 898 899 +---+---------------------------+-----------------------------------------------------------------------------+ 900 | # | Stage name | Description | 901 | | | | 902 +===+===========================+=============================================================================+ 903 | 0 | Prefetch packet meta-data | #. Select next two packets from the burst of input packets. | 904 | | | | 905 | | | #. Prefetch packet meta-data containing the key and key signature. | 906 | | | | 907 +---+---------------------------+-----------------------------------------------------------------------------+ 908 | 1 | Prefetch table bucket | #. Read the key signature from the packet meta-data (for extendable bucket | 909 | | | hash tables) or read the key from the packet meta-data and compute key | 910 | | | signature (for LRU tables). | 911 | | | | 912 | | | #. Identify the bucket ID using the key signature. | 913 | | | | 914 | | | #. Prefetch the bucket. | 915 | | | | 916 +---+---------------------------+-----------------------------------------------------------------------------+ 917 | 2 | Prefetch table data | #. Read the bucket. | 918 | | | | 919 | | | #. Compare all 4 bucket keys against the input key. | 920 | | | | 921 | | | #. Report input key as lookup hit only when a match is identified (more | 922 | | | than one key match is not possible) | 923 | | | | 924 | | | #. For LRU tables only, use branchless logic to update the bucket LRU list | 925 | | | (the current key becomes the new MRU) only on lookup hit. | 926 | | | | 927 | | | #. Prefetch the key value (key data) associated with the matched key (to | 928 | | | avoid branches, this is done on both lookup hit and miss). | 929 | | | | 930 +---+---------------------------+-----------------------------------------------------------------------------+ 931 932Additional notes: 933 934#. The pipelined version of the bucket search algorithm is executed only if there are at least 5 packets in the burst of input packets. 935 If there are less than 5 packets in the burst of input packets, a non-optimized implementation of the bucket search algorithm is executed. 936 937#. For extendable bucket hash tables only, 938 once the pipelined version of the bucket search algorithm has been executed for all the packets in the burst of input packets, 939 the non-optimized implementation of the bucket search algorithm is also executed for any packets that did not produce a lookup hit, 940 but have the bucket in extended state. 941 As result of executing the non-optimized version, some of these packets may produce a lookup hit or lookup miss. 942 This does not impact the performance of the key lookup operation, 943 as the probability of having the bucket in extended state is relatively small. 944 945Pipeline Library Design 946----------------------- 947 948A pipeline is defined by: 949 950#. The set of input ports; 951 952#. The set of output ports; 953 954#. The set of tables; 955 956#. The set of actions. 957 958The input ports are connected with the output ports through tree-like topologies of interconnected tables. 959The table entries contain the actions defining the operations to be executed on the input packets and the packet flow within the pipeline. 960 961Connectivity of Ports and Tables 962~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 963 964To avoid any dependencies on the order in which pipeline elements are created, 965the connectivity of pipeline elements is defined after all the pipeline input ports, 966output ports and tables have been created. 967 968General connectivity rules: 969 970#. Each input port is connected to a single table. No input port should be left unconnected; 971 972#. The table connectivity to other tables or to output ports is regulated by the next hop actions of each table entry and the default table entry. 973 The table connectivity is fluid, as the table entries and the default table entry can be updated during run-time. 974 975 * A table can have multiple entries (including the default entry) connected to the same output port. 976 A table can have different entries connected to different output ports. 977 Different tables can have entries (including default table entry) connected to the same output port. 978 979 * A table can have multiple entries (including the default entry) connected to another table, 980 in which case all these entries have to point to the same table. 981 This constraint is enforced by the API and prevents tree-like topologies from being created (allowing table chaining only), 982 with the purpose of simplifying the implementation of the pipeline run-time execution engine. 983 984Port Actions 985~~~~~~~~~~~~ 986 987Port Action Handler 988^^^^^^^^^^^^^^^^^^^ 989 990An action handler can be assigned to each input/output port to define actions to be executed on each input packet that is received by the port. 991Defining the action handler for a specific input/output port is optional (i.e. the action handler can be disabled). 992 993For input ports, the action handler is executed after RX function. For output ports, the action handler is executed before the TX function. 994 995The action handler can decide to drop packets. 996 997Table Actions 998~~~~~~~~~~~~~ 999 1000Table Action Handler 1001^^^^^^^^^^^^^^^^^^^^ 1002 1003An action handler to be executed on each input packet can be assigned to each table. 1004Defining the action handler for a specific table is optional (i.e. the action handler can be disabled). 1005 1006The action handler is executed after the table lookup operation is performed and the table entry associated with each input packet is identified. 1007The action handler can only handle the user-defined actions, while the reserved actions (e.g. the next hop actions) are handled by the Packet Framework. 1008The action handler can decide to drop the input packet. 1009 1010Reserved Actions 1011^^^^^^^^^^^^^^^^ 1012 1013The reserved actions are handled directly by the Packet Framework without the user being able to change their meaning 1014through the table action handler configuration. 1015A special category of the reserved actions is represented by the next hop actions, which regulate the packet flow between input ports, 1016tables and output ports through the pipeline. 1017:numref:`table_qos_33` lists the next hop actions. 1018 1019.. _table_qos_33: 1020 1021.. table:: Next Hop Actions (Reserved) 1022 1023 +---+---------------------+-----------------------------------------------------------------------------------+ 1024 | # | Next hop action | Description | 1025 | | | | 1026 +===+=====================+===================================================================================+ 1027 | 1 | Drop | Drop the current packet. | 1028 | | | | 1029 +---+---------------------+-----------------------------------------------------------------------------------+ 1030 | 2 | Send to output port | Send the current packet to specified output port. The output port ID is metadata | 1031 | | | stored in the same table entry. | 1032 | | | | 1033 +---+---------------------+-----------------------------------------------------------------------------------+ 1034 | 3 | Send to table | Send the current packet to specified table. The table ID is metadata stored in | 1035 | | | the same table entry. | 1036 | | | | 1037 +---+---------------------+-----------------------------------------------------------------------------------+ 1038 1039User Actions 1040^^^^^^^^^^^^ 1041 1042For each table, the meaning of user actions is defined through the configuration of the table action handler. 1043Different tables can be configured with different action handlers, therefore the meaning of the user actions 1044and their associated meta-data is private to each table. 1045Within the same table, all the table entries (including the table default entry) share the same definition 1046for the user actions and their associated meta-data, 1047with each table entry having its own set of enabled user actions and its own copy of the action meta-data. 1048:numref:`table_qos_34` contains a non-exhaustive list of user action examples. 1049 1050.. _table_qos_34: 1051 1052.. table:: User Action Examples 1053 1054 +---+-----------------------------------+---------------------------------------------------------------------+ 1055 | # | User action | Description | 1056 | | | | 1057 +===+===================================+=====================================================================+ 1058 | 1 | Metering | Per flow traffic metering using the srTCM and trTCM algorithms. | 1059 | | | | 1060 +---+-----------------------------------+---------------------------------------------------------------------+ 1061 | 2 | Statistics | Update the statistics counters maintained per flow. | 1062 | | | | 1063 +---+-----------------------------------+---------------------------------------------------------------------+ 1064 | 3 | App ID | Per flow state machine fed by variable length sequence of packets | 1065 | | | at the flow initialization with the purpose of identifying the | 1066 | | | traffic type and application. | 1067 | | | | 1068 +---+-----------------------------------+---------------------------------------------------------------------+ 1069 | 4 | Push/pop labels | Push/pop VLAN/MPLS labels to/from the current packet. | 1070 | | | | 1071 +---+-----------------------------------+---------------------------------------------------------------------+ 1072 | 5 | Network Address Translation (NAT) | Translate between the internal (LAN) and external (WAN) IP | 1073 | | | destination/source address and/or L4 protocol destination/source | 1074 | | | port. | 1075 | | | | 1076 +---+-----------------------------------+---------------------------------------------------------------------+ 1077 | 6 | TTL update | Decrement IP TTL and, in case of IPv4 packets, update the IP | 1078 | | | checksum. | 1079 | | | | 1080 +---+-----------------------------------+---------------------------------------------------------------------+ 1081 1082Multicore Scaling 1083----------------- 1084 1085A complex application is typically split across multiple cores, with cores communicating through SW queues. 1086There is usually a performance limit on the number of table lookups 1087and actions that can be fitted on the same CPU core due to HW constraints like: 1088available CPU cycles, cache memory size, cache transfer BW, memory transfer BW, etc. 1089 1090As the application is split across multiple CPU cores, the Packet Framework facilitates the creation of several pipelines, 1091the assignment of each such pipeline to a different CPU core 1092and the interconnection of all CPU core-level pipelines into a single application-level complex pipeline. 1093For example, if CPU core A is assigned to run pipeline P1 and CPU core B pipeline P2, 1094then the interconnection of P1 with P2 could be achieved by having the same set of SW queues act like output ports 1095for P1 and input ports for P2. 1096 1097This approach enables the application development using the pipeline, run-to-completion (clustered) or hybrid (mixed) models. 1098 1099It is allowed for the same core to run several pipelines, but it is not allowed for several cores to run the same pipeline. 1100 1101Shared Data Structures 1102~~~~~~~~~~~~~~~~~~~~~~ 1103 1104The threads performing table lookup are actually table writers rather than just readers. 1105Even if the specific table lookup algorithm is thread-safe for multiple readers 1106(e. g. read-only access of the search algorithm data structures is enough to conduct the lookup operation), 1107once the table entry for the current packet is identified, the thread is typically expected to update the action meta-data stored in the table entry 1108(e.g. increment the counter tracking the number of packets that hit this table entry), and thus modify the table entry. 1109During the time this thread is accessing this table entry (either writing or reading; duration is application specific), 1110for data consistency reasons, no other threads (threads performing table lookup or entry add/delete operations) are allowed to modify this table entry. 1111 1112Mechanisms to share the same table between multiple threads: 1113 1114#. **Multiple writer threads.** 1115 Threads need to use synchronization primitives like semaphores (distinct semaphore per table entry) or atomic instructions. 1116 The cost of semaphores is usually high, even when the semaphore is free. 1117 The cost of atomic instructions is normally higher than the cost of regular instructions. 1118 1119#. **Multiple writer threads, with single thread performing table lookup operations and multiple threads performing table entry add/delete operations.** 1120 The threads performing table entry add/delete operations send table update requests to the reader (typically through message passing queues), 1121 which does the actual table updates and then sends the response back to the request initiator. 1122 1123#. **Single writer thread performing table entry add/delete operations and multiple reader threads that perform table lookup operations with read-only access to the table entries.** 1124 The reader threads use the main table copy while the writer is updating the mirror copy. 1125 Once the writer update is done, the writer can signal to the readers and busy wait until all readers swaps between the mirror copy (which now becomes the main copy) and 1126 the mirror copy (which now becomes the main copy). 1127 1128Interfacing with Accelerators 1129----------------------------- 1130 1131The presence of accelerators is usually detected during the initialization phase by inspecting the HW devices that are part of the system (e.g. by PCI bus enumeration). 1132Typical devices with acceleration capabilities are: 1133 1134* Inline accelerators: NICs, switches, FPGAs, etc; 1135 1136* Look-aside accelerators: chipsets, FPGAs, etc. 1137 1138Usually, to support a specific functional block, specific implementation of Packet Framework tables and/or ports and/or actions has to be provided for each accelerator, 1139with all the implementations sharing the same API: pure SW implementation (no acceleration), implementation using accelerator A, implementation using accelerator B, etc. 1140The selection between these implementations could be done at build time or at run-time (recommended), based on which accelerators are present in the system, 1141with no application changes required. 1142