Design Your Private Cloud With Hybrid in Mind

Polls and inquiries show a growing interest in hybrid cloud computing for infrastructure as a service (IaaS). While technologies and providers for the hybrid cloud model are very nascent, private cloud architectures today should be designed with hybrid in mind. Managing this evolution is part of the overall transformation toward "hybrid IT."

• Most private and community cloud services will evolve over time to a hybrid model.
• A cloud service that spans both private and public cloud implementations, or both on-premises private and off-premises private or public cloud implementations, is a hybrid cloud service.
• Hybrid cloud connections can be static (done at provisioning time) or dynamic (rebalancing constantly); static connections will be the most common.
• Technologies, vendors and business models are still nascent, but emerging to enable hybrid cloud services.

• Design private cloud deployments with interoperability and future hybrid cloud computing in mind.
• Choose vendors and technologies for private cloud carefully; there is a significant variety of vendor strategies on how to connect to cloud providers and which providers they enable.
• Especially in larger enterprises, consider the evolution to hybrid cloud computing as part of a broader strategy to position IT as the broker for a broad mix of IT services delivered in many different ways – hybrid IT.

Private cloud computing is a hot trend in 2012, but enterprises choosing vendors and building architectures need to design private clouds with a longer-term strategy in mind – enabling interoperability and hybrid cloud computing within the broader context of transforming IT toward hybrid IT (becoming the broker of IT services for the enterprise).

A private cloud service is defined by privacy, not location, ownership or management (see Figure 1). It is a service that is either used by a single user (typically an enterprise or agency) or an implementation that is unshared (typically delivered by a public cloud provider, as a fenced-off virtual private cloud offering). It can be on-premises or off-premises, customer- or provider-managed, and customer- or provider-owned. Conversely, a public cloud service is delivered by an external service provider, it is open to any user and the implementation is shared. A unique offering between the two is a community cloud service, which is like a private cloud service in that the users are limited to a defined set of users, but also like a public cloud service in that the implementation is shared with more than one user (typically, multiple enterprises or agencies) in a cooperative manner.

Private, Community and Public Cloud Services

Source: Gartner (February 2012)

Private and community cloud services are appealing to enterprises because they offer the necessary security and privacy, protect proprietary algorithms and data, maintain regulatory compliance and ensure service levels. The downside of on-premises private and community clouds is usually less elasticity and scaling – requiring more excess stand-by capacity, and therefore potentially higher capital costs (and risks). However, service provider offerings of hosted private, virtual private, public cloud services and blended offerings will mature. Also, technologies to bridge on-premises and off-premises cloud service providers are already emerging, and will continue to mature – both for IaaS (e.g., VMware vCloud Director) and platform as a service (e.g., Microsoft System Center 2012 App Controller with Azure). While private and community cloud services will be popular first-instance offerings of cloud services, they will evolve into hybrid cloud models over time.

There are many ways that an IT service can be considered "hybrid": a mix of physical and virtual resources, a mix of multiple providers, a mix of proprietary and public resources, a mix of private and public networking, a mix of cloud with noncloud services and resources, a mix of shared and nonshared equipment and a mix of on-premises with off-premises services. In addition, service providers that offer cloud services are also beginning to offer "blended hosting" services that include cloud services and more traditional (noncloud) hosting. For private cloud services, three important forms of hybrid are:

• On-premises private clouds combined with off-premises private clouds
• On-premises private clouds combined with off-premises public clouds
• Off-premises private clouds combined with off-premises public clouds

Hybrid cloud services like these horizontally span two complete implementations. A single service request could be deployed in either implementation, moved from one to the other, or can horizontally grow between the two implementations (also called cloudbursting or overdrafting). The primary benefit is flexibility of deployment, managed security and elasticity. This is contrasted with a cloud service that relies completely on another cloud service in a supply chain where the service is incomplete without the connection between the multiple implementations. The primary benefit of a cloud supply chain is separation of concerns (e.g., multiple cloud providers each doing part of the overall service).

The connection between the cloud services in a hybrid cloud can be more static, or more dynamic and pervasive. A static connection implies that a service instance is deployed to either the private cloud or to a secondary cloud service at provisioning time – perhaps based on policy, on current private cloud utilization or both. A static connection could also rebalance cloud placement of service instances when cloud service capacity requires it – for example, certain service instances could be moved to the secondary cloud service during peak loads, thus requiring more primary private cloud resources. Static connections will be much more prevalent, especially over the next few years as hybrid cloud computing moves from concept to more mainstream.

A dynamic connection would imply rapid scaling and rebalancing between cloud providers, eliminating the boundaries between the two, and perhaps even having service instances span both provider footprints. A dynamic connection will usually require significant application work to enable cross-cloud federation, significant work on gateways between two providers, and will be hampered by latency and speed-of-light constraints. While powerful in concept, dynamic rebalancing and cloud federation will remain a niche trend.

There will be several methods that enable federation between two cloud providers in a hybrid cloud service (and boundaries between them may blur over time):

• Direct connections will use the cloud service's APIs "directly" – for example, coded into an application.
• External cloud connectors enable interoperability to a cloud service provider (gateway), or talk to corresponding cloud connectors at the cloud service (bridge).
• Cloud service brokers can manage the role of external cloud connectors and manage communication with multiple cloud service providers. Cloud service brokers will commonly be third parties, especially for midmarket enterprises – in essence, system integrators for cloud services, often providing value-added services. In large enterprises, IT will often evolve to take on the role of cloud services brokerage – a part of IT's transition to managing hybrid IT.

Automating the process of rebalancing and sourcing decisions is the job of a service governor, or cloud orchestrator. Cloud management platforms (CMPs) are emerging that provide the technologies for orchestration, connection and brokering – but it is a very nascent market.

Although there have been few actual deployments of hybrid cloud services so far, market interest is very high. For virtualization-based IaaS, there is a strong interest in evolving to a hybrid model as services and technologies mature. During Gartner's December 2011 Data Center Conference, a road map describing stages of IaaS evolution, starting with virtualization, was presented. The path from virtualization to private to hybrid will be a common one (see Figure 2); however, enterprises could also start by leveraging public cloud services first and enable hybrid later.

The Virtualization Road Map Through Private Cloud Computing

Source: Gartner (February 2012)

Attendees were then asked the question, "By 2015, how would you describe your virtualization progress?" (see Figure 3).

Gartner Data Center Conference Poll (n = 104)

Source: Gartner (February 2012)

Forty-five percent said that they would want to be managing both their on-premises and off-premises virtual machines centrally. In conversations with attendees, we heard of significant plans to examine CMPs to enable a private cloud soon, and orchestration and hybrid connectivity in the future. Interest in hybrid cloud computing is high.

Cloud computing services and technologies continue to evolve. According to Gartner clients, the trend toward private cloud deployments is rapidly growing in 2012. CMPs are rapidly evolving, but the dozens of CMPs today will shake out significantly over the next two years through failures and acquisitions. Gartner recommends that private cloud deployments be designed with a longer-term strategy in mind – specifically, private cloud services should be designed to enable hybrid cloud computing in the future. Many CMP vendors have early hybrid connectivity technologies, but there is significant variety in the market in terms of how they interoperate, which service providers they enable and how the hybrid services can be managed. Build a CMP architecture based on immediate private cloud requirements, but ensuring future hybrid expansion.

The big picture is that large-enterprise IT is evolving from a host of equipment to a broker of IT-based services – whether they are traditional services, public cloud, private cloud, hybrid cloud or some blend. Gartner calls this new model hybrid IT. Midmarket and smaller enterprises will likely rely more on external cloud service brokers over time (and, in fact, until those brokers emerge, smaller enterprises won't move far into leveraging public cloud services). The evolution from private cloud deployments to hybrid to possibly even public for some services will be an opportunity for enterprises to decide where responsibility for service delivery lies, regardless of sourcing – within the enterprise IT department or with external brokers.

Most private cloud and community cloud services will evolve over time to a hybrid model – at least for service placement, provisioning, and, occasionally, movement between providers. The market for CMPs is immature and exploding, but the market will shake out due to failures and acquisitions during the next two years. Enterprises should design private cloud architectures with hybrid and provider interoperability in mind. Before deploying CMPs, understand what provider architectures are being enabled and which ones are not. Longer term, the evolution to hybrid cloud computing will be part of the overall transition of larger enterprise IT organizations to hybrid IT becoming the broker of a variety of services and sourcing for the enterprise.