Revisiting the Roots of Software-Defined Infrastructure

The Intel® Developer Forum in San Francisco is a big show covering a lot of topics.  But for me, IDF13 has been about extending the concepts of software-based control to all corners of the datacenter. 

This tone was set by Diane Bryant, senior vice president and general manager of the Datacenter and Connected Systems Group, when she introduced Intel’s software-defined infrastructure strategy during her IDF13 press briefing.  The SDI strategy is designed to help data centers adapt on the fly to changing workloads.

The implication is that more data center systems will become “software-defined,” so I thought it would be useful to reflect back on the first software-defined movement – software defined networking (SDN) and the Intel® Open Network Platform (Intel® ONP) strategy.

What is SDN?

SDN emerged as a response to growing complexity in today’s data center networks. Data center administrators are increasingly using server and network infrastructure virtualization to help make the most of their infrastructure and to meet the dramatic growth in data on a constrained budget.

SDN provides a way to help these administrators manage the added complexity that comes with increasing numbers of virtual and physical servers and networks.  SDN replaces the traditional IP addresses that are used to locate servers and direct data flows.  IP routers and switches were designed to manage static networks, but modern networks feature virtual device mobility and multi-tenancy, which slows IP routing table convergence.

SDN is based on a centralized controller that separates and abstracts the network’s control information from the underlying data forwarding infrastructure. This creates a tremendously dynamic and flexible infrastructure that can be easily automated and managed.

The controller is a key element of this and manages the flow tables that instruct switches how to handle traffic within the network.  The controller takes information from the control plane to update the flow table and communicates with the switch so that it can forward the data to the proper outbound port.

With intelligence centralized in the controller, the switches themselves can be optimized for performance and latency. Additionally, network applications – such as security, or load balancing – can be run on virtual servers attached to the controller.  Intel’s Open Network Platform provides the switches, processors and software that OEMs need to make the most of SDN. 

One key element of Intel ONP is the Intel® Open Network Platform Switch Reference Design (Intel® ONP Switch Reference Design), which provides 48x10GbE and 4x40GbE connections using the Intel® Ethernet Switch FM6000 Series silicon. This customer evaluation board also comes with a two-core Intel processor and the Intel® Communications Chipset 89xx Series, which provides, among other things, CPU off-load capability features such as encryption and decryption for security.  Several OEM customers will be showing systems based on the Intel at the show.

If you’d like the details about Open Network Platform, take a look at the Intel ONP press release that we issued in April, or watch this video on the topic.  Also, here’s a link to more information on SDI and other news from IDF13.