Hybrid Cloud Scenarios: Own the Base, Rent the Spike


The hybrid cloud is the perfect solution for organizations that crave the business agility, scalability and OPEX cost model of the public cloud, but want consistent performance and control of security, compliance and/or long-term costs. In our work with Equinix customers, we’ve seen these concerns come up again and again. Over time, we’ve developed five hybrid cloud scenarios that, from our experience, address customer concerns while leveraging the best aspects of the cloud.

It’s all about deciding what to put in the public cloud and what to run within your company’s infrastructure. There are four issues that drive this decision: cost, performance, security and compliance. The scenario you choose depends on the importance of each to your organization’s business and the parts of the infrastructure-data, network and application-they apply to.

High Performance, Low Cost

The first scenario, which we call “own the base and rent the spike,” is foundational and very much about cost and performance. A typical use case is a large e-commerce site that needs to support occasional online sales spikes, such as Cyber Monday online sales, without any performance loss. A public cloud Infrastructure-as-a-Service (IaaS) environment provides the scalability, agility and fast ramp-up to support these huge spikes. The performance scalability is primarily achieved by horizontal scaling of the application tier. IaaS is pay per use, so the organization only pays for the infrastructure it rents for the time it’s being rented.

Harnessing IaaS makes a lot of sense, as it’s not cost-efficient for a retailer to purchase, own and manage ALL the servers and other infrastructure necessary to support a Cyber Monday spike, given much of it will sit unused for the rest of the year. You can also determine which public service is best for the types of workloads you are running.

So that’s renting the spike. What about owning the base?

First, determine the amount of infrastructure needed to securely support the workload in a steady state for most of the year and then purchase and run that infrastructure internally. That’s clear for an existing application, but how do you determine your steady state for a new application? I would recommend using a method called “fail fast.” For example, start by running your entire workload in the public cloud until the steady state becomes clear. Then build the capacity you need to support that steady state in your private cloud and rent the rest of the capacity, as required, in public cloud. This works especially well for new business initiatives whose success is not completely assured. You risk nothing up front, but when success is validated, you have an understanding of the most cost- and performance-efficient way to split the workload across public and private boundaries.

The alternative, starting the initiative in the private cloud, risks both poor performance and overbuying private cloud infrastructure. Once you know how to split your workloads, containerize them using Docker to abstract the application from the underlying IaaS provider. Once containerized, use a multi/hybrid cloud management platform like Cisco InterCloud, Rightscale, Elastibox, Chef, or Puppet Labs to distribute the workload across public and private clouds automatically.

Go Direct and Private

For the fastest ramp-up, the least risk and the most predictable performance, go for private connectivity to your cloud provider, rather than risking the performance jitters of a public Internet connection. For latency sensitive workloads, locate your private infrastructure close to your public cloud environment to reduce latency and increase bandwidth for the maximum performance. Interestingly, the closer you are to the cloud provider, the cheaper the data transfer cost. Hence, don’t be penny wise and pound foolish here.

In addition, you can safely connect to public IaaS providers more securely with direct connections instead of going over the public Internet. This allows you to protect your crown jewels (i.e., classified data and assets) for meeting compliance and regulatory requirements without compromising performance.

Gaining the Best of Both Worlds

The benefit of the own the base, rent the spike scenario is it ensures maximum cost-efficiency, performance, security and reliability, with the lowest risk of lost revenues and customers. In essence it’s the best of both worlds, which, of course, is what the hybrid cloud is all about.

Learn more about how to avoid hybrid cloud gotchas.

Read the rest of the Hybrid Cloud Scenarios series:

Control the Data

Disaster Recovery

Control the Network

Control the Application