35. Test Pipeline Application

The Test Pipeline application illustrates the use of the DPDK Packet Framework tool suite. Its purpose is to demonstrate the performance of single-table DPDK pipelines.

35.1. Overview

The application uses three CPU cores:

Core A (“RX core”) receives traffic from the NIC ports and feeds core B with traffic through SW queues.
Core B (“Pipeline core”) implements a single-table DPDK pipeline whose type is selectable through specific command line parameter. Core B receives traffic from core A through software queues, processes it according to the actions configured in the table entries that are hit by the input packets and feeds it to core C through another set of software queues.
Core C (“TX core”) receives traffic from core B through software queues and sends it to the NIC ports for transmission.

Fig. 35.1 Test Pipeline Application

35.2. Compiling the Application

Go to the app/test directory:

export RTE_SDK=/path/to/rte_sdk
cd ${RTE_SDK}/app/test/test-pipeline

Set the target (a default target is used if not specified):
```
export RTE_TARGET=x86_64-native-linuxapp-gcc
```
Build the application:
```
make
```

35.3. Running the Application

35.3.1. Application Command Line

The application execution command line is:

./test-pipeline [EAL options] -- -p PORTMASK --TABLE_TYPE

The -c or -l EAL CPU coremask/corelist option has to contain exactly 3 CPU cores. The first CPU core in the core mask is assigned for core A, the second for core B and the third for core C.

The PORTMASK parameter must contain 2 or 4 ports.

35.3.2. Table Types and Behavior

Table 35.1 describes the table types used and how they are populated.

The hash tables are pre-populated with 16 million keys. For hash tables, the following parameters can be selected:

Configurable key size implementation or fixed (specialized) key size implementation (e.g. hash-8-ext or hash-spec-8-ext). The key size specialized implementations are expected to provide better performance for 8-byte and 16-byte key sizes, while the key-size-non-specialized implementation is expected to provide better performance for larger key sizes;
Key size (e.g. hash-spec-8-ext or hash-spec-16-ext). The available options are 8, 16 and 32 bytes;
Table type (e.g. hash-spec-16-ext or hash-spec-16-lru). The available options are ext (extendable bucket) or lru (least recently used).

Table 35.1 Table Types
#	TABLE_TYPE	Description of Core B Table	Pre-added Table Entries
1	none	Core B is not implementing a DPDK pipeline. Core B is implementing a pass-through from its input set of software queues to its output set of software queues.	N/A
2	stub	Stub table. Core B is implementing the same pass-through functionality as described for the “none” option by using the DPDK Packet Framework by using one stub table for each input NIC port.	N/A
3	hash-[spec]-8-lru	LRU hash table with 8-byte key size and 16 million entries.	16 million entries are successfully added to the hash table with the following key format: [4-byte index, 4 bytes of 0] The action configured for all table entries is “Sendto output port”, with the output port index uniformly distributed for the range of output ports. The default table rule (used in the case of a lookup miss) is to drop the packet. At run time, core A is creating the following lookup key and storing it into the packet meta data for core B to use for table lookup: [destination IPv4 address, 4 bytes of 0]
4	hash-[spec]-8-ext	Extendable bucket hash table with 8-byte key size and 16 million entries.	Same as hash-[spec]-8-lru table entries, above.
5	hash-[spec]-16-lru	LRU hash table with 16-byte key size and 16 million entries.	16 million entries are successfully added to the hash table with the following key format: [4-byte index, 12 bytes of 0] The action configured for all table entries is “Send to output port”, with the output port index uniformly distributed for the range of output ports. The default table rule (used in the case of a lookup miss) is to drop the packet. At run time, core A is creating the following lookup key and storing it into the packet meta data for core B to use for table lookup: [destination IPv4 address, 12 bytes of 0]
6	hash-[spec]-16-ext	Extendable bucket hash table with 16-byte key size and 16 million entries.	Same as hash-[spec]-16-lru table entries, above.
7	hash-[spec]-32-lru	LRU hash table with 32-byte key size and 16 million entries.	16 million entries are successfully added to the hash table with the following key format: [4-byte index, 28 bytes of 0]. The action configured for all table entries is “Send to output port”, with the output port index uniformly distributed for the range of output ports. The default table rule (used in the case of a lookup miss) is to drop the packet. At run time, core A is creating the following lookup key and storing it into the packet meta data for Lpmcore B to use for table lookup: [destination IPv4 address, 28 bytes of 0]
8	hash-[spec]-32-ext	Extendable bucket hash table with 32-byte key size and 16 million entries.	Same as hash-[spec]-32-lru table entries, above.
9	lpm	Longest Prefix Match (LPM) IPv4 table.	In the case of two ports, two routes are added to the table: [0.0.0.0/9 => send to output port 0] [0.128.0.0/9 => send to output port 1] In case of four ports, four entries are added to the table: [0.0.0.0/10 => send to output port 0] [0.64.0.0/10 => send to output port 1] [0.128.0.0/10 => send to output port 2] [0.192.0.0/10 => send to output port 3] The default table rule (used in the case of a lookup miss) is to drop the packet. At run time, core A is storing the IPv4 destination within the packet meta data to be later used by core B as the lookup key.
10	acl	Access Control List (ACL) table	In the case of two ports, two ACL rules are added to the table: [priority = 0 (highest), IPv4 source = ANY, IPv4 destination = 0.0.0.0/9, L4 protocol = ANY, TCP source port = ANY, TCP destination port = ANY => send to output port 0] [priority = 0 (highest), IPv4 source = ANY, IPv4 destination = 0.128.0.0/9, L4 protocol = ANY, TCP source port = ANY, TCP destination port = ANY => send to output port 0]. The default table rule (used in the case of a lookup miss) is to drop the packet.

35.3.3. Input Traffic

Regardless of the table type used for the core B pipeline, the same input traffic can be used to hit all table entries with uniform distribution, which results in uniform distribution of packets sent out on the set of output NIC ports. The profile for input traffic is TCP/IPv4 packets with:

destination IP address as A.B.C.D with A fixed to 0 and B, C,D random
source IP address fixed to 0.0.0.0
destination TCP port fixed to 0
source TCP port fixed to 0