It’s taken as a given that as networks become more software defined, they will become more intelligent as well. After all, once all this dynamic configuring and reconfiguring starts to happen, even the most adept network administrator will be hard-pressed to stay on top of things.
But this kind of intelligent management is not built into SDN. It has to be implemented through advanced hardware and/or software stacks and will likely require a fair amount of customization to suit the specific needs of the enterprise and the vagaries of legacy data environments.
On the hardware side, Juniper recently released the EX4600 switch as part of its High-IQ networking strategy. Not only does the device offer a whopping 1.44 Tbps throughput over 72 1 GBE ports, 72 10 GbE ports or 12 40 GbE ports in a single rack unit, it also holds an advanced analytics module that can be used to optimize network performance according to application needs. In addition, it supports the company’s Virtual Chassis architecture, which allows up to 10 switches to be managed as a single device.
Cisco is on a similar path with its Application Centric Infrastructure (ACI) but has taken the additional step of driving some of the intelligence into applications themselves. The company’s DevNet program is intended to foster a community of app developers who will tailor their offerings to Cisco APIs. In this way, both the network and the application can provide a two-way communications channel in order to provision the proper resources to meet I/O, workload and other connectivity requirements.
But with more enterprise infrastructure making its way to the cloud, intelligence in the data center or even the campus LAN is no longer enough. The wide area network (WAN) needs some lovin’, too. Silver Peak has a new WAN fabric called Unity that gathers information about cloud services and traffic conditions – what it terms “Internet weather” – in order to provide secure, optimized connections for SaaS and IaaS services. With the cloud taking on greater responsibility for higher-order enterprise functions, managers will need visibility into the entire data path, even those parts that fall outside the traditional LAN or WAN. The Unity Cloud Intelligence service provides a range of Web optimization functions, such as adaptive forward error correction and application prioritization, to help ensure low-loss, low-latency connectivity.
At the same time, an intelligent network will provide a higher level of security for the entire data environment. APCON’s Paul Ginn noted recently that hackers are increasingly turning their attention toward network vulnerabilities to slip through firewalls and other defenses, and detection is getting more difficult as enterprise data loads increase. An intelligent network can house tools like the Span Port Analyzer inline, rather than on the switch, allowing it to monitor more of the traffic flow without packet loss while eliminating the port contention that arises with the addition of every new monitoring tool. It also becomes easier to implement functions like packet slicing to help reduce overall data loads.
Some people consider network intelligence to be simply a more advanced form of automation. This is only partially true, however, as it also incorporates disciplines like visibility, monitoring, governance and policy development, application and data awareness, and, yes, human oversight.
The intelligent network can do a lot of things that today’s networks can’t, but one thing it can’t do is think for itself.
Photo courtesy of Shutterstock.