Quantcast
Channel: Intel Developer Zone Articles
Viewing all articles
Browse latest Browse all 3384

SDN, NFV, DPDK and the Open Networking Platform - FAQ

$
0
0

Frequently Asked Questions on SDN, NFV, DPDK and the Intel® Open Network Platform Server

Why are SDN and NFV so important?

Software Defined Networking (SDN) and Network Functions Virtualization (NFV) are emerging as an alternative to traditional network design because they address many of the drawbacks of traditional hardware based networks.  SDN and NFV offer four primary benefits:

  • faster deployment of new services (in terms of both time to market and quote to cash)
  • more flexibility in design of the network,
  • reduced OPEX costs, primarily through automation
  • reduced hardware costs in moving away from specialized equipment

While either SDN or NFV can be implemented independently, there are benefits in combining these technologies. NFV provides the virtualization architecture and SDN provides the APIs and control protocols. SDN provides a centralized controller (separated from the forwarding plane) and a network view that is programmable by external applications. OpenFlow is a communication protocol between the control and forwarding layers.

What is the effect on my network infrastructure?

As shown by the Intel® Open Network Platform Server, the Intel® ONP Server Reference Architecture provides hardware that is optimized for a comprehensive stack of the latest open ­source software releases. A virtualized network can be deployed using standard off the shelf server platforms, ranging from Intel® Atom to Intel® Xeon® Processors.

What are all these Open names?

The network and virtualization infrastructure tools define an information model, a set of APIs, and control protocols such as OpenFlow, Netconf, and TR-69. At Intel, we are investing in our ecosystem partners and developer programs to develop SDN and NFV solutions through open source projects.

OpenStack* (Juno*) provides the framework to create/manage VMs (virtual machines). VMs are the base OS’s for the virtualized functions. VMs can have multiple virtual Network Interfaces.
OpenStack Neutron* is the networking component that abstracts Linux network configurations using a common API wrapped around the network function solutions (Open vSwitch, VLANs, iptables/netfilter, etc).
OpenDaylight* (Helium) provides the code and architecture for virtualizing the network controller (control plane functions for configuration, monitoring, and management)
Open vSwitch* (OVS) is a production quality, multilayer, virtual network switch. (OVS can be a node connected to an OpenDaylight controller)
Data Plane Development Kit (DPDK) is a set of data plane libraries and NIC drivers that provide a programming framework for fast, high speed network packet processing on general purpose processors.

What is the value of the openness of networking solutions?

Open source software and open hardware allow Linux servers to become switches and other networking ‘appliances’ while providing high-performance routing optimizations. By using Linux for the control plane and offloading all packet processing to Intel ONP hardware, applications can now obtain knowledge about network behavior and provide a rapid response to their customers. The capability to program the network processor ASICs allows developers the flexibility to redirect flows from the hardware processing to their software or another middle box service to tailor packet processing needs on demand.

So what does Intel make?

Intel (including Wind River) are major contributors to the DPDK and to Linux. Intel recently merged the Intel® DPDK vSwitch into the Open vSwitch main branch so that Neutron can use Intel’s accelerated packet processing while avoiding proprietary plugins. By building the switching logic on top of the Intel DPDK library, there is a significant boost to the packet switching throughput which can be integrated in both the host and guests of the OpenStack network compute nodes.

The Intel DPDK also adds samples of L3 forwarding, load balancing, and timers, all of which can help reduce development time. It also exposes resources as multiple virtual functions making them available to multiple VMs and available to speed up inter-VM communication.

Additionally Intel is prototyping Open NFV (OPNFV) concepts using the OpenDaylight platform to leverage Intel’s network performance enhancements.

Intel® QuickAssist Acceleration Technology provides accelerator services (encryption, compression and algorithm offload) for up to 14 separate virtualized instantiations. Intel QuickAssist Technology is available on the Intel® EP80579 Integrated Processor, the Intel® Xeon® Processor E5-2600 and E5-2400 Series, and Intel® Core™, Intel® Pentium® and Intel® Celeron® Processors with Intel® Communications Chipset 89xx Series.

Intel is also providing the Intel® Open Network Platform Server. The Intel® ONP Server Reference Architecture provides hardware that is optimized for a comprehensive stack of the latest open ­source software releases, as a validated template to enable rapid development. The Reference Architecture includes architecture specs, test reports, scripts for optimization and support for Intel network interfaces from 1 GbE to 40Gbe (FTXL710-AM2 4x10GbE )

What SDN/NFV applications are available?

To bypass the bottlenecks within standard hypervisors (virtual switch and unoptimized vNIC Linux drivers), 6WINDGate leverages the Intel® DPDK to offload packets from the guest’s OS. 6WIND has also extended the Intel® DPDK with the support of virtual NICs (including, e1000).

Many more solution partners listed at  https://networkbuilders.intel.com/solutionslibrary

What might be an obstacle to implementing SDN and/or NFV?

The back office (OSS/BSS) is a possible barrier to deploying NFV and SDN. Intel is making a significant investment in defining application touch points to Operators BSS/OSS systems by developing a better understanding where the operators back office software needs to migrate to and how components need to interact when implementing NFV and SDN-based services.

Other challenges exist around Security and scale. The improved efficiency and flexibility of centralized control can come with a cost and may not fit all use cases. The network needs to be failure resilient and have elastic compute, storage, and network resource allocation. Especially for the network, this requires diagnostic and troubleshooting tools that combined reporting, monitoring,  and analysis with the ability to provide fine grain traffic control to inspect, redirect, transform, and ensure quality of service per port, per user, and per service type.

Where can I find more information? 

 


Intel maintains several sites including: 

Additional Resources

About the Authors

Thomas Kee is Intel’s Networking Community Manager and Networking Developer Evangelist. Thomas is a software engineer with over 20 years of experience developing routing and switching products and testing networking solutions.

Colleen Culbertson is a Platform Application Engineer with Intel Developer Relations Division - Scale Enabling.


Viewing all articles
Browse latest Browse all 3384

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>