Everyone's always talking about the wonders of cloud computing and how cloud technologies are advancing to provide end users with access to unlimited applications, data, and services. Yet there's a single bottleneck that can dramatically reduce the effectiveness of any cloud architecture: the network. Many enterprise organizations leapt into cloud architectures without fully considering the limitations that legacy networks place on overall application performance.
The days of simply throwing more bandwidth at network problems are long behind us. This is especially true as it relates to hybrid and public cloud deployments. Instead, a new approach to network optimization that focuses on the applications themselves is in order. In this article, I'll look at ways to modify the network that will help optimize cloud application and data access.
Easily the most popular method these days to alleviate network congestion to public and hybrid cloud resources is to bolster the internet and WAN edge. WAN optimization techniques such as compression, localized content caching, and manipulating TCP window sizing have been go-to methods for speeding up remote cloud access.
But SD-WAN technologies have proven their worth. A software-defined WAN essentially provides intelligent routing capabilities using two or more paths to remote destinations like the cloud. Path selection is based on constantly calculated variables such as available link bandwidth, lowest latency, and least amount of packet loss. The technology helps connect remote corporate locations to the cloud.
In the case of remote users connecting to cloud services from home offices or the road, network architects also need to consider a different approach. Before cloud computing took off, apps and data resided inside private data centers or colocations. Remote users commonly have used VPN technologies to securely connect to the main corporate office.
However, now that applications and data are spread among various cloud providers, it no longer makes sense to terminate users' remote-access VPN tunnels at the corporate headquarters only to hairpin back out to the internet or across expensive WAN links. Instead, network designers should work to streamline data flows to make the fewest number of network hops as possible. This usually means setting up streamlined, yet secure access directly to public cloud resources over the internet using IPsec VPN, SSL VPN, or secure web-based access using HTTPS.
Additionally, if you have end users spread around the globe, it would be wise to develop a globally distributed cloud architecture for applications that are latency sensitive. Globally distributed clouds speed up client-server access to the various regions they are deployed. That way, users can access resources that are located closer to them physically from a geographical perspective.
Today, network engineers need deep insight into how critical applications perform end-to-end. Not only is it important to understand how data flows operate between the server and end users, but also between the server and other distributed resources that may reside in different areas of a cloud, or in completely different clouds.
Having a clear understanding of data flows using modern network performance management (NPM) tools allows network engineers to optimize the placement of resources, so they are free from bottlenecks or congestion. It also paints a picture of what quality of service (QoS) identification, marking, and policy implementation an organization should enforce for the proper prioritization of business-critical apps.
While all the network optimization techniques described above will help improve cloud access, the ultimate goal for network professionals should be a complete, end-to-end software-driven network intelligence architecture. Unfortunately, this has been proven to be easier said than done from both an implementation and cost perspective. That's why we're seeing enterprise companies narrow their focus on solving cloud-access optimization problems within specific areas of the network like the WAN and data center. But eventually, the goal should be to deploy optimization policies across the entire network using a software-defined framework.