Last week at the Cloud Connect 2012 conference in Santa Clara, I was sharing a panel with industry colleagues representing the most prominent vendors in the application optimization, cloud infrastructure and network acceleration space . The topic was “hitching your wagon to the cloud”, discussing the importance of the network, particularly WAN, to make your cloud deployments successful. Folks talked about interesting concepts like “stateless branch office”, and “nirvana of Internet as WAN” before someone in the audience retorted, “I need solutions that help me deal with reality!”
The reality is that as enterprises adopt the cloud, IT teams, particularly networking, are expected to make everything work magically and deliver the best user experience without sacrificing security. In my last post, “Why hybrid clouds look like my grandma’s network”, I highlighted networking requirements on the cloud side to make deployments successful. Today, let’s focus on what’s needed on the branch or user side of the equation. It is indeed possible to deliver a stellar cloud experience to even the most far flung branches and users on the move if IT teams consider WAN obstacles and avoid the following pitfalls :
- Cloud is about the Internet, so WAN does not matter
- I can secure Cloud in the same way as I secure other enterprise applications
- Cloud is about centralization so can “survive” with dumb infrastructure at the branch
In this blog, let’s talk about the fallacy of “WAN does not matter.” In future blogs, I’ll cover the other pitfalls. A common misperception is that Cloud, especially SaaS applications like Salesforce.com, are accessible via the Internet and don’t need rethinking of WAN strategy. Adding bandwidth to remote sites should be good enough to provide a good user experience, right? Not necessarily. The reality is that the majority of enterprises still backhaul Internet traffic over private WAN, T1/T3 or MPLS, to central sites for Internet handoff. Just adding bandwidth is not only expensive, but also has the counter effect of adding more latency as traffic from the branch user to SaaS application causes hair-pinning via the central site.
What you need is to re-architect your WAN into a “hybrid WAN” with direct Internet access from the branch that intelligently uses your private WAN for intra-enterprise traffic. To do this, your network at the branch needs to be smart enough to:
- Split traffic between private and public networks based on policies
- Have a unified QoS framework to make sure critical Internet traffic does not get clobbered by someone watching a casual YouTube video
- Build failover capabilities to 3G/4G wireless networks for cases where the Internet connection goes down
Thoughts? Other pitfalls, you may have experienced?