In early 2009, when I first started working with enterprises thinking about building private clouds and with communication service providers (CSPs) thinking about building public clouds, a lot of the clients’ technical focus was on what I would consider basic capabilities.
It was typical to see self-service provisioning of instances based on a single virtual machine with variable operating system, memory and disk configurations. More advanced clients were looking for limited allocation-based showbackand usage monitoring. Larger enterprises and ambitious CSPs soon started thinking about building heterogeneous clouds, for example running Dev/Test x86 workloads on an open-source hypervisor like kvm, production x86 workloads on VMWare and with ERP applications running on a Power-based system.
By late 2009 and early 2010 I started seeing enterprises thinking about placing some workloads on private clouds and some workloads on public clouds. For example, one client was moving an order processing application onto their new private cloud, but their sales team had been using Salesforce.com for a year already. CIOs started to understand that Hybrid Cloud Computing had evolved from an idea to a reality, and they needed to understand their role in delivering, managing, securing, and providing business resilience for these new services and service delivery technologies.
The US National Institute for Standards and Technology (NIST) defines Hybrid Cloud as “a composition of two or more distinct cloud infrastructures (private, community, or public) that remain unique entities, but are bound together by standardized or proprietary technology that enables data and application portability.”
This is a well-thought-out description, and I’d like to discuss how I have seen clients across various industries start exploring the Hybrid model in interesting ways. I’ll group these into two macro use cases that I’ll call “Governed Cloud Management” and “Hybrid Cloud Scaling” for simplicity.
First, I’d like to discuss “Governed Cloud Management” as exemplified by a case I heard in early 2011 from a client in the US in the Healthcare industry. The CIO of this company, which shall nameless, managed a large and complex IT environment, with massive regulatory and compliance requirements. The IT Department was understaffed and over worked, and essentially ran from one emergency to the other, while providing less than stellar service to the business. The CIO was in talks with multiple vendors about building a private cloud infrastructure using some new and some existing hardware. His goals were straightforward: increase agility and reduce costs. He wanted a simple self-service catalog, with a single hypervisor and some basic monitoring.
Then he found out that a clinical trial group, impatient with the IT department, had started using their corporate credit cards to order vast amounts of virtual machines from a public cloud provider to run their analytics. “I could go to jail for this,” he told me, “this could be your grandmother’s blood pressure data out there.” He quickly realized that he needed policy-driven, governed access to all the various cloud-delivered services used in his enterprise. For example while he did not want clinical trial analysts to use unsecured public cloud services, it could be appropriate for marketing teams.
This “Governed Cloud Management” requirement also extended to the private cloud infrastructure. The CIO realized he needed a layer of policy-driven management that could control which cloud services and which cloud resources were available to which users. And he realized that the business might even expect him to provide these services backed up, secured, and with a certain level of uptime guarantee. Additionally, his governance solution would need to scale to future cloud architectures. Without an open management layer, and clear standardization between applications, providers, hypervisors and application structures, the challenge was significantly more complex that he originally thought.
Clients started talking about “Hybrid Cloud Scaling” use cases around the same time, in late 2010 and early 2011. One of the earliest cases I recall was a European online electronics retailer. Their IT infrastructure hosted their product catalog and online ordering systems. Every few months, they would run a special offer, with heavy advertising, offering for example a free Blue-ray player with all orders above 400 Euros during a particular 48-hour period. Of course the traffic to their website would spike during these times, but they did not want to buy new IT infrastructure to support these bursts of demand. The company looked to a public cloud provider to provide peak capacity to support this extra demand. The challenge was developing and implementing an architecture to support this “outbound spillover” capability.
To effectively use public infrastructure, the client’s architecture monitors their order processing application in real time to detect lengthy response times. When pre-defined response time thresholds are reached and sustained, the Cloud orchestration layer automatically provisions a new webserver and order processing instances at the public cloud provider. These public instances connect the webserver to their in-house database, and client activity is load-balanced between the on- and off-premise instances. This provides a seamless customer experience, while successfully offloading peak workloads to their external cloud provider. Additionally, with integrated monitoring of both their off-premise and on-premise resources, the client built an integrated dashboard that gave fully visibility into quality of service across both public and private infrastructure.
This concept extends as well to “inbound spillover” for internal applications as well. Consider peak internal workloads, for example quarter end financial processing or year-end benefits enrollment traffic to a company’s intranet. For these peak workload periods, I am seeing more and more enterprises seeking pay-as-you-go virtual infrastructure and even middleware that are securely integrated with the client’s own network infrastructure. Advanced clients are starting to see this in many ways as a convergence between traditional IT outsourcing providers and cloud service providers.
Given the surge in client interest I’ve seen lately in adopting and exploiting the technology and business benefits of hybrid cloud computing, I’m predicting 2012 will be the Year of Hybrid Cloud. Client-driven requirements, in addition to the ones I’ve mentioned, will grow more aggressive and will link closer to measures of business agility and cost efficiency. The best approach for both clients and technology vendors will be to drive greater openness and connectivity to harness public and private infrastructure to implement a Hybrid Cloud. The recent announcement of the OASIS-led Topology and Orchestration Specification for Cloud Applications (TOSCA) comes at a perfect time to help make Hybrid Cloud a key deployment model for business agility.
Source:http://www.wired.com/cloudline/2012/01/2012-the-year-of-hybrid-cloud/
