UX in network apps: how not to fail
UX in network apps: how not to fail
Header Background Image

Data Plane Development Kit (DPDK)

CodiLime Glossary

Some common networking terms clearly explained

Glossary /D /

Data Plane Development Kit (DPDK)

DPDK (Data Plane Development Kit) is a set of libraries for implementing user space drivers for Network Interface Controllers. DPDK provides you with a framework and common API for high-speed networking applications and allows for achieving a fast packet processing pipeline.

Data Plane Development Kit has been designed to provide a simple, yet complete framework for fast packet processing in networking apps. It’s an open-source software project managed by the Linux Foundation, with a large development community that is working on its improvement.

How does DPDK work?

DPDK provides a set of data plane libraries and network interface controller polling-mode drivers. DPDK makes communication between NIC and application in user space possible without kernel involvement. Thus, networking applications can run much faster. DPDK has five core components: the Environment Abstraction Layer (the EAL – for access to low-level resources, like memory space), the MBUF (a data structure that carries network packets as messages), the MEMPOOL (the library for creating allocated memory packets) the RING (the library that manages log-less messages sent between threads, cores, etc.), and the TIMER (for asynchronistic callback functions).

Why use DPDK?

DPDK is particularly useful when you want to accelerate packet processing on multiple-core CPUs. It allows developers and software engineers to develop high-speed data packet networking applications faster. DPDK enables more efficient computing than the standard interrupt processing that is available in the kernel. It actually eliminates the need for interrupts. Instead of using them, the Data Plane Development Kit uses a PMD (a poll mode driver) that continuously monitors Network Interface Controllers activity. Finally, DPDK works smoothly with cloud databases.

Read more: