markrubens - Fotolia


Before you deploy OpenStack, address cost, hybrid cloud issues

OpenStack deployments are complex and involve power, hybrid cloud, performance, update and cost concerns. Research each carefully to avoid these OpenStack challenges.

Although OpenStack provides many of the tools and technologies necessary for a private cloud, the path to private cloud can still be long and difficult. Business and IT leaders must confront significant hurdles to successfully deploy an OpenStack private cloud.

Be aware of issues when IT deploys OpenStack

One central challenge with OpenStack is its complexity. There are numerous modules designed to provide a wealth of important services, such as Nova for compute, Heat for orchestration and Cinder for block storage. But at its heart, OpenStack is a middleware framework, not an everyday application or workload that you can easily install. OpenStack can successfully deliver an effective and flexible cloud operating environment for the enterprise, but the investment itself can be costly and time-consuming.

Traditional IT silos don't work. The whole purpose of OpenStack and cloud is to create a truly software-driven data center by pooling and provisioning resources across multiple domains, such as compute, storage, network, security and identity. To configure and deploy OpenStack, you must have a solid understanding of many traditional silos and be able to work seamlessly across those silos to set up and manage the OpenStack environment.

OpenStack hybrid cloud integration

When IT deploys OpenStack to construct a private cloud in the data center, it's often just the beginning. Many organizations eventually extend the private cloud into public cloud to form a ubiquitous hybrid cloud environment.

A hybrid cloud offers several important benefits, including resource and workload fluidity, scalability and resilience, along with an even more tantalizing potential for multicloud hybrid environments, which can completely eliminate the risks associated with public cloud vendor lock-in.

However, hybrid cloud integration is easier said than done. OpenStack operates through a series of APIs, and those APIs don't readily interoperate with major public cloud APIs right out of the box. Initiatives to establish a common set of APIs are still in their infancy and have been slowed by a lack of industry standards. Even if such integration were readily available today, there are other concerns for enterprise hybrid clouds, such as monitoring and reporting on the hybrid infrastructure, resources, workloads and performance. Hybrid integration is possible, but it can require yet another substantial investment in a software layer or third-party service.

For example, services from companies such as Platform9 offer OpenStack drivers to control and manage resources on Amazon Web Services (AWS). This allows core OpenStack modules, like Nova, Glance, Neutron and Cinder, to integrate with AWS and support management of AWS endpoints using OpenStack. Microsoft's Azure Stack brings Azure services and capabilities to the on-premises data center in the form of an appliance. However, Azure Stack is not an open source platform, and users are committed to an Azure-based cloud.

OpenStack performance and resilience

OpenStack provides a native orchestration engine in the Heat module, which is also compatible with the AWS CloudFormation template format. But OpenStack also supports popular third-party orchestration tools for VMs and containers. For example, the vRealize Orchestrator plug-in for OpenStack allows administrators to orchestrate and automate VMware workflows through OpenStack. Container users can also use container orchestration engines, like Kubernetes.

In spite of this versatility, after your deploy OpenStack, carefully evaluate the critical security, scalability and resilience features that are often associated with orchestration and automation. You may encounter difficulties creating robust workloads -- especially with complex, scalable workloads. And once you deploy workloads, you may face challenges in performance monitoring, service-level assurance and optimizing infrastructure usage. Deploy OpenStack in proof-of-principle projects to determine precisely how it fits business needs. You may need to deploy additional tools and policies to meet any shortcomings in these areas.

OpenStack costs can be high

Open source software is free, and those upfront cost savings are often attractive to organizations. The problem is that free software is never actually free. Simply getting OpenStack to work can represent a significant investment in IT and developer staff time -- especially when the project winds up running late. Most of this cost is startup, but there is often additional investment in testing, deploying and configuring new OpenStack releases.

Training can become an indirect OpenStack cost. IT and developer staff may not have the requisite skill sets needed to tackle an OpenStack deployment. You may need to find more OpenStack-savvy staff to handle the job, spend the money to train up existing staff as Certified OpenStack Administrators, hire consultants to jump-start the work or some combination of these tactics.

Consider the implications of OpenStack support. Organizations can certainly adopt a canned OpenStack distribution and associated support from vendors like Red Hat or Rackspace. As open source software acquired directly, however, there is no official support. If you choose to deploy OpenStack, assemble a suite of support resources to address inevitable questions or to resolve problems. Some resources are free, while other resources will incur added costs. For example, the OpenStack site lists numerous how-to guides and other documentation, as well as an extensive marketplace with connections to various service organizations that offer OpenStack expertise for deployment, support and training.

OpenStack future

OpenStack is supported and moved forward by an active contributor community -- the latest release of OpenStack Ocata involved work by 1,925 individuals from 265 organizations. There are several vendor-based distributions and support for OpenStack available, including Canonical OpenStack, Red Hat OpenStack Platform and many others. Even code-savvy adopters can modify APIs in-house to tweak integration, performance and resource utilization for the individual data center environment.

But this blessing can also be a curse for an enterprise that expects clear and direct product leadership. Enterprise executives draw confidence from the well-defined direction of a single organization before making such a substantial investment in the data center and IT operations for the business. It's not just a matter of software cost or software development but in underlying hardware, OSes, IT staff, training and the surrounding business implications of policy, process and procedure involving OpenStack services.

The challenge for OpenStack is that its feature set and future roadmap are both community-driven also. The OpenStack roadmap currently extends to the next three releases beyond Ocata: Pike, Queens and Rocky. Adopters must either be willing to go with the community or manage and maintain OpenStack software themselves.

Next Steps

Enterprise OpenStack prepares to move in

How can OpenStack get ready for the enterprise?

OpenStack still awaits full enterprise adoption acceptance

Dig Deeper on Enterprise data storage strategies