The challenges faced by network operators are constantly growing. Just as they seem to find their way around a new technology or concept, another one comes into force, requiring them to rack their brains yet again. The only way they can keep up with the pace of technological change is by staying abreast with the latest trends in network technology. Network Functions Virtualization (NFV) is one such trend that is helping network operators create a more consolidated, agile, and cost-effective network infrastructure. It is enabled by Intel’s Data Plane Development Kit or DPDK.
The Need for NFV
As competition (and complexity) in the telecom industry intensifies, network operators often find themselves struggling to improve network efficiency while optimizing costs. NFV, through decoupling of firewalls, intrusion systems, and encryption from dedicated hardware onto virtual servers, allows operators to accelerate service deployment in a scalable and cost-effective manner.
This brand-new way of designing, deploying, and managing networking services makes it easy for operators to keep up with changing trends and demands while saving up humongous costs in the installation of new equipment. Using NFV, operators can
- Greatly bring down CapEx while also optimizing OpEx and making it more predictable
- Accelerate time-to-market through the quick provisioning of new services
- Improve business agility and flexibility by scaling resources as per the demand
Core Components of DPDK
As network operators increasingly look to leverage NFVs to drive greater efficiencies and deploy new revenue-generating services rapidly, DPDK constitutes a crucial element of NFV ecosystems – especially as the bandwidth requirements from Intel x86 architectures are constantly growing.
DPDK, an open-source software developed by Intel, offers a set of libraries that help network operators accelerate packet processing on multiple-core CPUs. It facilitates the quicker expansion of high-speed data packet networking applications while paving the way for more efficient computing.
Since it works well with the cloud, it provides a simple, comprehensive framework for fast packet processing in data plane applications. Using DPDK, telecom companies can more easily receive and send packets within the minimum number of CPU cycles, develop fast packet capture algorithms, as well as run third-party fast path stacks.
Here are the 5 core components of DPDK:
- Timer: The RTE timer library provides a timer service to RTE Data Plane execution units, allowing for asynchronous execution of callback functions. Using this library, operators can add, delete, or restart a timer.
- Memory pool: The RTE memory pool library allows operators to handle a pool of objects while allowing for bulk enqueue/dequeue. This memory pool uses a ring to store free objects and provides additional services such as per-core object cache to ensure objects are spread equally on all RAM channels.
- EAL: The environment abstraction layer provides a generic interface to enable access to low-level resources such as hardware and memory space. It hides the environment specifics from the applications and libraries and while providing mechanisms for assigning execution units to specific cores, tracing and debugging functions as well as determine if a particular feature is supported at runtime.
- Ring: The RTE Ring Manager library provides a lockless, multi-producer, multi-consumer FIFO API that makes it easy to store objects in a table. It is easier to implement, and it can also adapt to bulk operations. It can be used as a general communication mechanism between cores or execution blocks connected on a logical core.
- MBUF: The network packet buffer management library delivers the ability to create and destroy buffers created or used by the DPDK application to store message buffers. Using APIs, operators can allocate or free mbufs as well as manipulate packet buffers which are used to carry network packets.
Tips for migrating to DPDK
For network operators looking to improve the efficiency of their growing networks, DPDK offers several frameworks that make moving to NFV or COTS-based hardware much easier. However, for a successful migration to DPDK, careful consideration at multiple levels is required:
- Make sure you have clarity on the expected throughput performance
- Determine if DPI is used – if yes, assess its impact on your operations
- Check how much application-level processing and data forwarding is expected
- Verify and confirm the future roadmap of the product and its alignment with your business
- If required, re-architect the forwarding plane to ensure it meets all the performance and functional requirements.
- Develop a Proof of Concept before completely migrating to DPDK to unearth key issues and plan for solutions
Partner with a DPDK expert
Partnering with DPDK experts like Benison who have extensive knowledge and experience in the development of NFVs and migration to COTS-based hardware is a great way to ensure successful implementation. Since experts have widespread experience working with major equipment vendors and helping them with the migration, they can help in understanding your motivation to move to COTS hardware, study your existing architecture, provide recommendations on the most suitable architecture, and build a robust product roadmap.