Data Plane Development Kit (DPDK) on Azure offers a faster user space packet processing framework for performance intensive applications that bypass the virtual machine’s kernel network stack.

Typical packet processing using the kernel network stack is interrupt driven. Each time the network interface receives incoming packets, there is a kernel interrupt to process the packet and context switch from kernel space to user space. DPDK eliminates context switching and the interrupt driven method in favor of a user space implementation using poll mode drivers for fast packet processing.

DPDK consists of set of user space libraries providing access to lower-level resources such as hardware, logical cores, memory management, and poll mode drivers for network interface cards.

DPDK can run in Azure virtual machines, supporting multiple operating system distributions. DPDK provides a key performance differentiation in driving network function virtualization implementations, in the form of network virtual appliances (NVA) such as a virtual router, firewall, VPN, load balancer, evolved packet core, and denial-of-service (DDoS) applications.

**Higher packets per second (PPS)**: Bypassing the kernel and taking control of packets in user space reduces the cycle count by eliminating context switch and improves the rate of packets processed per second in Azure Linux virtual machines.

| Linux OS | Kernel Version |

Refer to [Patches for building an Azure-tuned Linux kernel](https://github.com/microsoft/azure-linux-kernel) for any Linux kernel version not listed, or for further information, contact [azuredpdk@microsoft.com](mailto:azuredpdk@microsoft.com).

2. First build the default config with `make config T=x86_64-native-linuxapp-gcc`.

Run the following commands once, after rebooting:

* Check hugepages are reserved with `grep Huge /proc/meminfo`.

> There is a way to modify the grub file so that huge pages are reserved on boot by following the [instructions](http://dpdk.org/doc/guides/linux_gsg/sys_reqs.html#use-of-hugepages-in-the-linux-environment) for DPDK. The instruction is at the bottom of the page. When running in an Azure Linux virtual machine, modify files under /etc/config/grub.d instead, to reserve hugepages across reboots.

2. MAC & IP addresses: Use `ifconfig –a` to view the MAC and IP address of the network interfaces. The *VF* network interface and *NETVSC* network interface have the same MAC address, but only the *NETVSC* network interface has an IP address. VF interfaces are running as slave interfaces of NETVSC interfaces.

* Find out which PCI address to use for *VF* with `ethtool -i <vf interface name>`.

* Ensure that testpmd doesn’t accidentally take over the VF pci device for *eth0*, if *eth0* has accelerated networking enabled. If DPDK application has accidentally taken over the management network interface and causes loss of your SSH connection, use the serial console to kill DPDK application, or to stop or start the virtual machine.

DPDK applications must run over the failsafe PMD that is exposed in Azure. If the application runs directly over the VF PMD, it will not receive **all** packets destined to the VM, since some packets will show up over the synthetic interface. Running over the failsafe PMD guarantees that the application receives all packets destined to it and also ensures the application will continue to run in DPDK mode, even if the VF is revoked when the host is being serviced. For more information on failsafe PMD, refer to [Fail-safe poll mode driver library](http://doc.dpdk.org/guides/nics/fail_safe.html).

Use `sudo` before the *testpmd* command to run in root mode.

If running with more than 2 NICs, the `--vdev` argument follows this pattern: `net_vdev_netvsc<id>,iface=<vf’s pairing eth>`.

3. Once started, run `show port info all` to check port information. You should see one or two DPDK ports that are net_failsafe (not *net_mlx4*).

The previous commands start *testpmd* in interactive mode, which is recommended, to try out some testpmd commands.

When running the previous commands on a virtual machine, change *IP_SRC_ADDR* and *IP_DST_ADDR* in `app/test-pmd/txonly.c` to match the actual IP address of the virtual machines before you compile. Otherwise, the packets are dropped before reaching the receiver.

When running the previous commands on a virtual machine, change *IP_SRC_ADDR* and *IP_DST_ADDR* in `app/test-pmd/txonly.c` to match the actual IP address of the virtual machines before you compile. Otherwise, the packets are dropped before reaching the forwarder. You won’t be able to have a third machine receive forwarded traffic, because the *testpmd* forwarder doesn’t modify the layer 3 addresses, unless you make some code changes.