Evolution of Optimized Private Clouds
With the potential of private clouds now being realized by organizations of all sizes, IT is awakening to the importance of continuously improving the way it manages and delivers cloud-based services.
With the potential of private clouds now being realized by organizations of all sizes, IT is awakening to the importance of continuously improving the way it manages and delivers cloud-based services. With speed, agility and efficiency gains at the center of private cloud adoption, IT can now shift its focus to enhancing and improving upon what it has already built.
At the core of this improvement is a holistic and continual approach of streamlining service management processes, intelligently fulfilling provisioning requests based on real-time capacity information, and optimizing the configuration and performance of virtual data.
A private cloud should not be considered the final destination; it should be treated as a critical milestone in the ongoing process of delivering cost-effective and efficient services and infrastructure. I've heard quite a few organizations declare "mission accomplished" by delivering private clouds based on a rigid set of vendor-published architectural components, with little to no understanding of the importance to continuously evolve and improve.
What's after the private cloud? For some, it is hybrid and public clouds, but for all, the journey should begin with the optimization of private clouds.
With the industry eagerly embracing an elastic and hypervisor-agnostic data center, now is a great time to define a new set of goals, objectives and milestones. Whether you call it a refined approach to IT-as-a-Service, private cloud 2.0, just good IT, or consumer-driven IT, it leverages your existing investment in private clouds and takes an evolutionary approach to improving quality of service while lowering your total cost of delivery. It sounds simple, until you begin to realize what this means for your organization (see Table 1).
|Table 1. Characteristics of an Optimized Private Cloud
||Optimized Private Cloud
|Scope of private cloud
||Partially integrated IT service management and provisioning
• Fully integrated IT service management and automated provisioning
(governed by policies)
• IT costing integrated with service requests
||Primarily infrastructure focused
||Infrastructure and service focused
|Deployment, placement and decommissioning of workloads
||Requested by end user, determined and fulfilled by IT based on service tier and/or available capacity
||Consumer initiated, determined by cost models, service level agreements and application response times
|Management capabilities and processes
• Hypervisor-centric with limited ability to leverage external resources
• Hypervisor-agnostic with ability to leverage external resources
• Integrated IT costing to drive placement of workloads
• Processes support future interoperability with hybrid clouds
|Resource optimization and monitoring
• Manual and rarely performed
• Quotas and policies
• Integrated capacity and performance management capabilities to drive rightsizing
• Quotas, policies, monitoring and integrated chargeback
IT cost models
|Primarily show back
||Integrated IT costing and chargeback
Expanded scope of private clouds: With the exception of small and greenfield data centers, rarely will IT bite off organization-wide cloud adoption, and for good reason. Most have already succumbed to some type of large and expensive project that has never really produced any quantifiable results. Just as virtualization and private clouds provided pooling of resources, now is the time to take advantage of economies of scale and expand across organizational structures. With standardized processes and automated management capabilities, it should be easy to demonstrate how the private cloud has increased agility and improved customer service. Ensuring service level agreements (SLAs) and demonstrating improved delivery times is a sure way to expand your footprint and capitalize on even more cost savings.
Integrated service management: Arguably, there are many different forms and levels of IT service management (ITSM) in private clouds, but if your deployment processes are disconnected from service management, you are most likely experiencing longer approval and deployment times. By integrating IT service management with provisioning you will be able to:
- Eliminate manual, error-prone processes
- Measure service fulfillment times (from the initial request through deployment)
- Curb consumption by integrating IT costing and chargeback information into request forms and service catalog entries
- Leverage policies to govern automated provisioning
Service catalogs: Catalogs provide IT a way to publish a roles-based listing of services available to users. Optimized private clouds recognize that while consumers of IT require infrastructure, they are actually more focused on consuming applications and services. By focusing on the end-to-end delivering of the service stack, IT realizes the benefits of streamlined operations while masking the user from the dependencies, complexity, and nuances of multi-tiered applications.
Deployment, placement, and decommissioning of workloads: This has typically focused on placing workloads on a single hypervisor that is largely driven by policies or tiers of services requested (gold, silver, bronze). By evolving this deployment and placement model to include multiple hypervisors, organizations can now offer multiple tiers of services, with varying costs, SLAs, and ideally, application response times. Not all workloads are created equal, nor should they be treated equally. In an optimized private cloud, consumers will not only be given incentives to consume lower cost infrastructures, they will naturally select appropriate SLAs and response times for workloads in question. Why? Because they are paying for it. Once consumers pay for consumption, they often become good stewards and begin rightsizing, decommissioning and recycling. Why? Because there is financial incentive to do so.
Management capabilities and standardized processes: These processes are at the center of any private cloud implementation. If you have implemented processes or management solutions that are only designed for a single hypervisor, evaluate your long-term strategy. You definitely don't want to be locked into a single hypervisor, and certainly not by your management vendor. Ensure that existing management capabilities support and facilitate the scale-out of infrastructure within your physical data center, as well as scale-out to the safe and secure private cloud boundaries of your hosted infrastructure. The ability to leverage an internal cost model that supports the intelligent placement of workloads across multiple hypervisors with the framework to support interoperability between external hybrid and public clouds is equally as important.
Hypervisor support: This should be a key consideration in evolving your private cloud strategy. Be wary of any hypervisor management vendors that claim to support multiple hypervisors. In most cases, support means "migrate from." It is not in their best interests to be hypervisor-agnostic, and they are motivated by getting your workloads on their platforms and keeping them there. Leverage lower cost, non-VMware alternatives like Microsoft's Hyper-V, Citrix XenServer, and Kernel-based Virtual Machine (KVM) when it makes sense.
Resource optimization: This gives private clouds the ability to do more with less. While standardization of builds (CPU, memory, disk, network, OS configuration) is paramount in ensuring consistency across your data center, don't let it add unnecessary costs. Sure, your customers might be paying for it, but most would rather take excess capacity and reallocate it to more critical workloads. Regularly rightsizing (both upsizing and downsizing) VMs across your data center will help ensure the optimal use of resources while eliminating costs associated with over-provisioning. Augment rightsizing with regular monitoring for out-of-inventory, offline, or dormant VMs, and you can quickly reclaim 20 percent of your capacity on a continual basis.
For non-critical workloads like development, quality assurance, and test, quotas and policies usually suffice in preventing over-consumption. However, in production, while quotas and policies have their role, it should not be at the expense of application performance and response times. A better and more refined approach is achieved through resource metering and integrated chargeback, which allow the application owner or consumer to request a known configuration with the ability to scale up or down as needed. As the service scales up, so does the cost. As the service scales down, so does the cost. Integrate cost, SLA and application response models into your private cloud to drive financial accountability into workload placements and provide financial incentives to migrate non-critical workloads to lower cost hypervisor alternatives when it makes sense.
IT cost models: While showback -- showing costs to the customer -- has proven to be a viable approach in curbing consumption in private clouds, it does little to curb ongoing consumption and abuse. In most implementations of showback, the estimated cost of the service is shown to the user at the time of request. Once the request is fulfilled, showback often does little to address the bulk of perpetual abuse. By implementing a robust IT cost model that is integrated with the initial request, provisioning processes, and billing cycle, not only do you change behavior when requests are made, but you also modify behavior when services are provisioned. This provides incentives for good behavior and financial consequences for bad behavior, while establishing a delivery model that is shielded from political influence. Combined with SLAs and service response times, you now have a powerful delivery model that not only masks the underlying hypervisor, but also enables consumers to make the most cost-effective decisions.
In conclusion, transforming the way IT delivers services and infrastructure is a never-ending process that is only hindered by the imagination. Over the last couple of years, this transformation manifested itself in private cloud adoption and has quickly evolved to include a wide range of requirements including integrated IT costing and chargeback, multiple hypervisor support, interoperability with external clouds, service-centric service catalogs, and integrated service management. As the private cloud becomes known as just "the cloud," look for an exciting new list of evolutionary requirements to emerge.