For networking folks, the good old days are fading away.
Applications used to be easy to manage, at least compared to today. Traditional network architecture approaches align networks with the applications they're supporting. There are linear data flows, which lead to linear networking flows, and they evolve together. As the application grows, so does the network. These topologies are, relatively, easy to scale horizontally using tools like load balancers, and simple to monitor by tapping single points of traffic flow.
"Then, virtualization changed things," said Eric Hanselman, chief analyst at the 451 Group, who presented a discussion on how cloud is changing networking at Cloud Connect event in Chicago this week. The fundamental difference is that virtualization allows applications to be mobile now. "When you start to move workloads around, those data flows become much more complicated," Hanselman said. "Those traditional networking tiers start to come undone."
[MORE NETWORKING:10 Most powerful network management companies]
Now, cloud has introduced a whole new set of complexity beyond just virtualization. In a cloud environment, not only are virtual machines sliced up within a server, but they can be automatically provisioned and scaled. That requires even more network flexibility. The cloud brings other challenges too. Today, applications can run in a geographically dispersed setting all around the world. But when that happens, the underlying data that supports those apps still needs to be constantly updated and synchronized as well.
Customers may be used to replicating data for disaster recovery scenarios. But that active/backup model doesn't quite cut it in a cloud world. Geographically dispersed applications need to be synchronized, creating active-active scenarios across multiple sites. But that's tough to do.
Sometimes users are settling with not having the most consistent data across these applications, but instead have a system where data is mostly up to date in real time, and will eventually replicate across a distributed environment, Hanselman said. Hyperscale data centers use this philosophy: When a user updates Facebook, that update may not show up immediately across the entire globe. But, eventually it will work its way through the system. "Prepare to separate and distribute" your data, Hanselman recommends.
The cloud has introduced these new models for data to be distributed across the globe, and the networking needs to keep up. Cloud providers are trying to make these networking paradigms advance with the advent of cloud services. Amazon offers direct connect, which is a direct link between their data centers and collocation facilities operated by a variety of partners, like Equinix. Other companies are rolling out SDN-like qualities in their cloud, which gives customers the ability to spin up and down networks on demand.
[FROM SMALL TO BIG:5 tips for managing your cloud at scale]
The cloud introduces much more dynamic characteristics to IT, says Bernard Golden, director of the enterprise practice at Entratius, a cloud management platform that is owned by Dell. Users usually understand the changes that are needed in the network from a conceptual point of view to accommodate this, but they confront these challenges once their systems are implemented.
A key to relieve some of these issues, he said, is to install some sort of software layer between the network and these dynamic applications whether that be a SDN, or more palatable virtualized switches. "Essentially, you need to have smart software in the middle of the network now," said Golden, who's also author of the recent Amazon Web Services for Dummies book.
The takeaway, Hanselman said, is that virtualization, and especially cloud, significantly alter traditional networking approaches. If the network doesn't update with the adoption of new technologies like cloud computing, the entire system can be flawed.