It seems like a no-brainer that wide area network (WAN) technologies would be an easy fix for application and data performance problems in the cloud.
This is only true up to a point, however. As experience with the cloud grows, it is becoming clear that existing WAN solutions aimed at connecting the home office with various branches don’t provide the same benefits when applied to highly dynamic, massively scalable external infrastructure.
In the first place, says Aryaka Networks’ Ajit Gupta, traditional MPLS architectures that require WAN controllers on both sides of the link are both too expensive and too inflexible for the cloud. As the cloud environment scales upward, packet loss and latency variations become more frequent, leading to performance degradations that significantly inhibit the ability to maintain enterprise-class service. What’s needed is an intelligent point-of-presence (POP) architecture that incorporates numerous, low-cost VPN links rather than a single MPLS connection.
Of course, WAN acceleration is much easier to manage when dealing with a single cloud provider. Silver Peak Systems, for instance, has launched a new Virtual Acceleration Open Architecture (VXOA) system in conjunction with Amazon Web Services that is said to boost cloud access 100-fold, even when host and client are separated by thousands of miles. The system cuts network traffic in half and improves migration and recovery operations with a 20x gain in throughput. It also works across all major hypervisors and can be installed on legacy enterprise systems in a few minutes.
Other developers are turning toward new software-defined network (SDN) architectures to improve wide area connectivity — not just to the cloud but to all disparate entities. BTI Systems recently released the Intelligent Cloud Connect platform that employs label-switch routing (LSR), packet-optical transport, and various application server functions within an integrated SDN fabric. The idea is to identify and remove network choke points that arise between data repositories and end points using forwarding module and application module blade servers that support data rates from 10 to 100 Gbps. In this way, network optimization can be quickly integrated into new platforms as they are added to the network, while open APIs ensure that third-party SDN platforms are supported.
Still others are toying with Network-as-a-Service (NaaS) approaches. A start-up called Pertino recently came out of stealth mode with a NaaS solution aimed at providing enterprise-class networking for small and medium-sized enterprises. The system allows users to quickly establish scalable WAN services using a simple Internet connection and then shut the entire operation down when no longer needed. At the same time, it avoids the substantial up-front cost of establishing a dedicated WAN infrastructure, with professional rates as low as $10 per month per user.
Such a variety of approaches is evidence of the fact that networking is king when it comes to the cloud. With server and storage resources available in unlimited quantities, the movement of data is now the primary driver of enterprise infrastructure.
Traditional WAN technologies will likely remain the options of choice for dedicated data architectures, but the new world of the cloud will require a high degree of flexibility and unwavering demand for low costs.
The primary task right now is to figure out the best way to accomplish those goals.
