Mike Bushong is Senior Director of Strategic Marketing at Juniper Networks.
The future promised by multi-cloud is at odds with the current reality in enterprise IT. This means change is going to require more than an incremental continuation of current architectural and operational practices. Practically speaking, enterprises need to plan holistically as they not only shift applications to the cloud but to something that is more multi-cloud ready.
The Tenets of Multi-cloud
When planning for a multi-cloud future, it is important to lay out the basic tenets driving multi-cloud architectures:
Security: With data at the center of the IT universe, security is more than a bolt-on. It has to be a top-tier architectural consideration, especially when users and workloads are distributed.
Ubiquity: One of the central theses of multi-cloud is that applications and services need to be everywhere. Indeed, if experience is dependent on location, the full promise of cloud will go undelivered.
Reliability: Expectations of infrastructure are approaching those of public utilities—everything must be available all the time. Even small gaps in availability are intolerable, which means reliability must be guaranteed in a multi-cloud world.
Fungibility: To drive application and service ubiquity in a reliable way without breaking the bank, resources must be exchangeable. That is to say, workloads cannot be bound to specific resources such that it impedes availability.
These tenets demand different architectural priorities from what’s traditionally been considered as the foundation of enterprise design.
Historical Enterprise Priorities
For decades, enterprise architectures have been dictated by two primary considerations, cost and complexity, with the goal of keeping both minimal. In both cases, the prevailing strategy is often containment.
Cost is generally well understood. Traditionally, IT has been seen as a cost center providing service to the business. In the cost center model, IT is typically funded by a corporate tax, paid by the lines of business. As businesses seek to maximize their profits, there is intrinsic pressure to keep costs down.
Complexity is a bit more interesting. In the general sense, complexity is not something that can be controlled. This is due to it being a function of multiple variables, such as the number of users, devices, applications, and so on. In a given system of a given size and makeup, complexity will be fixed. For example, an environment running 15 years’ worth of accumulated technology will be more complex than a homogenous data center with common building blocks. No amount of abstraction removes the operational considerations required to manage a diverse infrastructure.
If complexity cannot be eliminated, then the strategy shifts to containment. The predominant mechanism here is to separate the infrastructure into well-defined domains surrounded by hardened boundaries. Each place in the infrastructure has its own devices managed by its own teams responsible for its own workflows to drive its own set of policies.
Multi-cloud Challenging Prevailing Approaches
This containment strategy was an absolute necessity to scale infrastructure to the levels of today. But as the world moves to multi-cloud, the model must fundamentally change.
For the multi-cloud tenets to be true, the boundaries between groups of resources must come down. Users need to have uniform experience regardless of where they are accessing workloads, and regardless of which resources are being used to service the workloads. Security and policy must be uniformly enforced. That simply doesn’t happen if there are hard boundaries that prohibit visibility, policy, and operational control at the boundaries between different domains.
Infrastructure Without Borders
This design goal, however, brings with it a set of technical issues that must be addressed:
End-to-end security policies: Policy and control cannot be domain-specific. If security is to be consistent, it has to be administered over-the-top. This places policy management requirements on all places in the infrastructure, from data center to branch.
End-to-end visibility: Policy is only part of the equation. For policy to be administered, systems and operators must have end-to-end visibility within the infrastructure. This is especially true as workloads require more coordination of resources in a multicloud architecture, meaning systems must be designed with telemetry and data modeling in mind.
End-to-end operational control: Ultimately, the workflows that administer these environments must extend beyond domain boundaries. If management becomes manual, the entire system breaks down, which makes automated control a must-have for multicloud environments. Accordingly, enterprises must decide on a control model, which will drive API requirements and data distribution mechanisms. For policy-related operations, there must be a software layer that translates application or user-side intent into device behavior, accounting for dynamic orchestration that is typical of cloud ecosystems.
Keeping Complexity in Check
If the premise of siloed design was to contain complexity, bringing everything back together is going to remove the safeguard that has made IT manageable until now. How do enterprises cope when they lose their best defense against a crippling foe?
As more companies migrate to multicloud, a new class of tools are emerging, with many of them meant to combat complexity. Whether it’s multi-domain orchestration or end-to-end visibility, these tools are quickly becoming the foundation for a new kind of architecture expressly built with multicloud in mind. The key will be seamlessly integrating these tools across the whole of the enterprise (data center, campus, branch, and public cloud), and also across the entire technology stack (from application layers down to transport).
More Than Just Technology
The biggest change in all of this, however, will prove to be cultural. There is a degree of muscle memory in how infrastructure is managed, and breaking design habits and operational practices is going to be difficult.
Large, black-out-the-sun type projects will be doomed to failure, not because of the products but because of the design philosophy. The best enterprises can do is to think holistically and think about where they want to be and take deliberate actions when expanding teams or purchasing technology that allow them to reach future objectives. Make sure that every decision helps unify the infrastructure in a bid to make it multi-cloud ready.
Opinions expressed in the article above do not necessarily reflect the opinions of Data Center Knowledge and Informa.
Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating.