Using Real-Time Programmable Analytics to Build Smarter and More Secure Networks

By Netronome | Mar 10, 2016

Lately, the server-based networking discussion has been heavily focused on performance gains associated with offload to SmartNICs such as Netronome's Agilio CX. This is an important topic and the discussion should continue, but I also think it’s important that we do not lose sight of another important aspect to offloading server-based networking, and that is the ability to drastically increase the networking capabilities possible at the edge of the network. With expanding CPU processing capacity combined with increasingly SmartNICs for networking, some amazing features can be implemented on the server that improve visibility, performance, security, and scalability.

This will be a topic for discussion at the Open Networking Summit (March 13-17, 2016) at the Santa Clara Convention Center. One of our lead researchers at Netronome, Nic Viljoen, will be a co-presenter with AT&T experts on a plenary session titled, “The Need for Complex Analytics from Forwarding Pipelines.” During this session, Nic will present an overview of how intelligent programmable server adapters can enable highly impactful networking features such as DDoS, elastic classification, and resource monitoring.

This new paradigm of intelligence is a result of the trend of shifting away from traditional middlebox appliances and specialized equipment to a specialization at the edge, enabled by software. This means there will be a greater number of applications and increased complexity of those applications that has not been observed in the past. This sophistication at the edge was not previously possible because edge servers were limited by CPU cycle capacity and networking bandwidth limitations at the core. However, those days are behind us with a new generation of SmartNICs and higher core-count x86 CPUs. Together, these technologies have drastically increased computation capacity; unlocking a level of analytics and monitoring that have not yet been observed. Operators can easily monitor VM performance, network traffic characteristics, and network threats. The real power of this technology lies in the ability to rapidly react to findings from increased analytics. This includes resource (re)allocation for VMs, rapid network and VM provisioning, and quick deployment of ACLs and stateful filtering.

Next week at ONS, Nic will discuss these topics in detail along with AT&T. If you will be attending ONS next week, I encourage you to check out the session at 2:40 p.m. on Tuesday in Grand Ballroom F and stop by Netronome’s booth 23 for a demonstration. Thanks and have a great week at ONS 2016!