While much of the networking industry today is focused on improving speeds and feeds inside the data center, we need to recognize the importance of improving the networks that connect enterprise data centers to each other, and to the public cloud. If the industry can deliver an elastic network with programmable performance, then the walls between data centers could effectively disappear.
Trying to overlay cloud services on the same pipe being used for best-effort internet is going to disappoint users, and limit cloud service adoption. Specifically, we need to add speed and intelligence to these networks, and several factors are driving this requirement. For example:
- Virtual machine (VM) transfers between data centers are increasingly common
- Virtual storage is no longer isolated to a single data center
- An increasing number of mission critical enterprise applications being deployed on VMs are moving to the cloud, driving the need for carrier-class network security and performance for workload balancing and reliability.
- Adoption of cloud-based infrastructure (IaaS) for workload mobility, collaboration and availability is creating more complicated topology deployments and opportunities for software defined networking.
Let’s imagine a company with a 200 Mbps data connection to the world, which needs to make a server platform change. To do so without shutting down the business in the process, IT staff would like to temporarily move the applications on this server to the public cloud. Let’s assume the total data to transfer would be about 10 terabytes to make this migration happen. However, transferring 10 TB of data over a typical 200 Mbps network connection would take nearly a week, assuming full bandwidth utilization, no re-transmissions and 80 percent utilization. Clearly this company is not going to be able to run this simple workload job over this network service.
This issue is quite debilitating for IT organizations, and is something service providers like Verizon have been hearing about from their enterprise customers. In fact, the company just opened a new innovation center dedicated to finding solutions that improve the integration of networks and data center infrastructure.
To make this work, enterprises require new cloud network connectivity options for efficient operations—an intelligent Network-as-a-Service model that uses software defined networking to dynamically provide performance as dictated by the application. In the cloud world, demands on capacity and connectivity are fluid — entirely dependent on businesses’ specific requirements at any given time. The network supporting this environment needs to be as elastic, programmable and, in a sense, “virtualized,” as storage and servers are today.
Using the example above for a 10 TB data transfer, an intelligent network could more easily accommodate these workloads by dynamically expanding to 5 Gbps and completing the job in less than 5 hours without requiring the VM applications to go offline. When the job is done, the network would immediately return to standard levels so that the premium bandwidth is billed only as used.
The fictional company finds this use of the premium network service worthwhile, quite simply, because it makes using the provider cloud practical for this particular workload. On-premise data center capital and operational cost becomes avoidable, replaceable with a time-limited -– and thus net-smaller –- IaaS “rental” expense.
Enterprises and service providers both benefit. Enterprises minimize permanent data center-related costs and reduce return-on-investment risks, while providers attract more workload and demand to their cloud services, which boosts their revenues. More, and more affordable, network when needed is essential to vreate these benefits.
The ability to respond to varying workload demands with performance generated on demand is a key benefit of an intelligent network for the cloud. In addition to dynamic bandwidth, this network must have higher availability, lower latency and greater reliability, as it would be designed for critical infrastructure services. Programmable interfaces into an open cloud networking framework might also be used to adjust for policies, authentication or network events.
This open, programmable networking model can be implemented as a cloud backbone or as a fully integrated cloud and network operation. A single vendor could provide a dynamic packet transport core and data center endpoints. Or multi-vendor switching and transport equipment can be used in the core with data center connect performance optimization and cloud operations at the end points.
Many vendors are adopting an open network philosophy and looking to implement interoperability by using new open protocol standards and application programming interfaces (APIs) with a virtualized network.
With this sort of network, the IT manager will have the freedom to consider resources outside the physical walls of his or her building to be natural extensions of an owned data center. In effect, that IT manager would now have a “data center without walls,” that provides the same user experience as a completely dedicated data center, but on a partially rented, and thus more economical, basis.
Rick Dodd is vice president of marketing at Ciena Corp.