Application-Aware Networks, or Network-Aware Applications?

Jim Poole

By Jim Poole

One of the highlights from day one (11th Oct) of Carrier Ethernet World Congress 2011 in Amsterdam was a panel discussion involving senior representatives from Colt, Cisco, Global Crossing, Ciena, and Brocade communications.

The debate was around how the Cloud and Ethernet are linked – and over the course of the session the focus tended to be on cloud application delivery over Ethernet.

As one might expect from a diverse group of panellists, there was agreement on some issues, and disagreement on others.

Panelists (from left to right) Nicolas Fischbach, Director, Network Architecture, Colt; Bart Van De Velde, Director Business Development, Cisco; Simon Parry, Specialist, Regional Portfolio Solutions, Ciena; Jeff Smith, Senior Director, Product Management, Global Crossing; Rajesh Dhople, Senior Product Manager, Brocade Communications

<<>>

Ciena’s Simon Parry set the debate in motion suggesting that “it’s too hard for an application developer to be thinking about the network, it’s the network that needs to think bigger.”

This point of view was echoed by Brocade’s Rajesh Dhople, who claimed that it is the network’s role to empower cloud applications, and so the industry needs to think in terms of “software-defined networking.”

However the counterpoint is best summarised by saying that making network hardware so ubiquitously high performing that it can’t cause a problem for cloud applications is simply not realistic. That would require a huge amount of capital investment from network companies, with the end result being their own networks reduced to commodity level -performance would be taken for granted, the only remaining differentiator would be price.

Perhaps this is a little like asking a turkey to vote for Christmas.

If a network were to invest heavily in this way, then it would have to restrict the benefits to its own added value cloud services to ensure a return on investment, rather than creating an equal benefit for all cloud players.

Cisco’s Bart Van De Velde made another valid counterpoint – that a race to the bottom only benefits everyone until the networks go wrong, because ultimately a lot of the current cost stems from ensuring reliability as well as performance, this is why the SLA exists, and it’s a reason why commoditisation has a limit: Some applications are just inherently more critical than others and therefore require more robust networks at a higher price.

In the end, the pragmatic view is that the ideal synergy of cloud and Ethernet will involve work on both sides.

Networks need to re-adjust their thinking, particular in terms of how they manage SLAs and test their own performance. What matters in cloud delivery is application performance, and in many circumstances this can be poor despite what appears on the surface to be a good network service.

But on the other side, networks can’t do it all alone – application developers and providers need to become more network savvy so that their applications perform better in the current environment and ultimately have an advantage over lesser performing applications.

Moreover, both parties should be looking to neutral data centers with rich ecosystems to help ensure that carrier and application are as tightly linked in the most efficient way possible.

While work can be done on both sides, it is ultimately the neutral data center where application and carrier can connect before making their journey to the end-user. Carriers should be locating in a neutral data center to expand the services available on the networks, while cloud application providers should be in a neutral data center to expand the connectivity options between themselves and their customers.