Campus Networks

Simon Pollard

Elastic Networks

by Simon Pollard ‎06-10-2014 12:27 AM - edited ‎06-10-2014 12:27 AM (1,144 Views)

One of the biggest challenges when designing any new system that requires an element of capacity planning – computing, networks, transportation, to name a few - is deciding what the primary criteria are; should you go for a big initial costly build-out and hope that all of the capacity is used? Or aim for something more flexible that can start smaller and be expanded as required to deliver greater capacity. Factors such as the need for up-front investment and sketchy predictions of future demand need to be considered and the result can sometimes be a difficult compromise.

 

When it comes to designing new networks the latest generation of Packet Processors deliver performance and scale at levels well beyond what we have seen in the past allowing over one hundred 10GbE line-rate interfaces to be packed into compact 1U form factor switches. These advances result in a new paradigm of price vs. performance in network switch design. However, things become really interesting if we can seamlessly connect these powerful fixed configuration devices together to create customized solutions that combine scalability and flexibility. Elastic networks are becoming a reality.

 

Scalability is used to describe a platform that has an architecture that can be expanded to meet greater demand, ideally that scalability should be linear so that the initial investment is not excessively large and growth costs are incremental. Flexibility means that different types of scalability can be can be mixed - do you want primarily 10GbE ports or more 40GbE ports - it’s the ability to choose without excessive compromises. Mid-range chassis based systems that provide 10GbE and 40GbE connectivity deliver non-linear scalability. Buying into a chassis is expensive for small deployments and in larger ones the concentration of capacity in one location results in the wrong kind of flexibility, sure there are lots of ports but they may not located where they are needed.

 

What is needed is an elastic solution. Elasticity is used to describe how well a solution architecture can adapt to varying demands that may not be linear. Brocade’s ICX7750, with its Distributed Chassis capability, gives network architects new options when building out the aggregation and core of their campus networks. And just like the line card options that are available for traditional chassis based systems the ICX7750 is available in a variety of formats; mixed 10G SFP+ and 40G ports, mixed 10G copper and 40G or all 40G so the required interface mixes can be combined in a single distributed system. Flexibility and scalability with the added benefits of a pay-as-you grow linear cost model and physical deployment options that put the ports where you need them means that the system becomes elastic. Want to put each switch into the top of a different rack or place them in different wiring closets? No problem, switches can be added, removed and physically placed wherever works best for the users and the business with no compromise in performance and with a linear cost model – yet each distributed chassis system can be managed as a single logical device. Welcome to the elastic network.

 

For the facts about elastic networking read the scale-out networks solution paper.