How to Think About Scalability for Your Enterprise Cloud Network
Building agility, flexibility and cost control into your application delivery environment
You want your enterprise cloud network to deliver all the benefits of a public cloud service—the elasticity to scale high performance on-demand, a pay-as-you-go cost structure, the power to affordably support an ever-expanding array of apps and customers. It’s hard to achieve that kind of scalability with a traditional fixed-capacity networking solution, and chassis-based solutions fall short on flexibility and cost. Only a software-defined network can deliver every kind of scalability you need.
Consider just a few of the many types of application delivery scenarios your cloud network needs to be able to handle:
- Steady and predictable rises in application traffic
- Temporary spikes in demand, like seasonal traffic that rises for a while, then subsides
- Abrupt, permanent increases in performance requirements—say, following a new business acquisition
- Inefficient or inadequate coverage by your application delivery controllers (ADCs) that you need to re-allocate to better support your tenants
To handle all the scenarios that matter, you have to make sure you’re thinking about scalability in every way that matters. Your cloud networking strategy should ensure scalability in three dimensions: up, out and in.
Rising traffic levels can be steady or sudden, permanent or temporary, predictable or not. However and whenever they arise, you have to be able to adjust your ADC performance to meet SLAs. This shouldn’t call for added hardware, disruptive forklift upgrades or costly over-provisioning. Instead, take a software-based approach with on-demand licensing that lets you increase performance quickly when you need to, then scale back down if and when the traffic eases up again.
Scaling up the performance of a single ADC appliance is one thing—but what about when you need to build out your capacity to support major changes in your business, like a merger than doubles the users for a critical application? The traditional approach would be either to use forklift ADC upgrades to boost your throughput, or segment your network into subsets served by ADCs paired for high availability.
What you really need, of course, is a way to use multiple ADCs—whether physical or virtual—to work transparently as a single logical ADC to deliver one or more app to every user. That way, you can continue to use your existing ADCs and add new appliances incrementally as needed to match network capabilities to your actual traffic requirements. It’s simpler to manage, more transparent to users, and ensures reliable performance without the idle resources of high-availability pairs.
People often think about scalability in terms of performance and capacity, but there’s another dimension that’s just as important: the ability to scale the number of apps or business units that you can support. Usually, you’d handle this by deploying any number of dedicated application delivery infrastructures—an approach that leads to all-too-familiar appliance sprawl and network complexity. Here’s a simpler way: using a virtualized architecture to support multiple tenants with a single physical device. After all, you’re already working to break down dedicated application silos to a more cost-effective shared infrastructure model—why should your ADCs be any different? With virtualized application delivery, you can achieve huge savings through consolidation while completely isolating individual application delivery instances, each with a full set of ADC capabilities.
The common theme running through each of these approaches is software—and that’s no coincidence. The essence of cloud computing is to use network functions virtualization to deliver resources in a more agile, scalable and cost-effective way. By building your enterprise cloud environment to allow software-based software scalability, you can keep pace with changing business conditions more effectively and achieve all the benefits of the cloud within your own organization.