How to Avoid the Bankrupt Private Cloud Part 1/2
Lessons learned from successful Virtual Data Center pioneers
This is the first in a three part blog series that releases a new white paper with the above title. Today’s initial installment introduces the paper and looks at Lessons # 1 and 2.
Check back tomorrow for Lessons 3 and 4.
What if you built a multi-million dollar virtualized private cloud, and all you had running on it was print servers?
According to leading analysts, the journey to “cloud computing” has already begun for most enterprises. It starts with server virtualization initiatives and ends in a model where IT services are delivered like a “utility,” and resource provisioning is dynamic in a “pay-as-you-go” model.
For most enterprise IT departments, however, this is uncharted territory. And without a road map developed by the pioneers, you may fall far short of your goals. Your new expensive infrastructure will still just be running print servers – and not business-critical applications. So how do you prevent this from happening?
Netuitive customers have built some of the world’s largest deployments of virtualized infrastructure, with ambitious plans for cloud computing. Many lessons were learned as they struggled to refine their architectures with the goal of successfully migrating Tier 1 applications to their private clouds – all while delivering even better, more proactive service management. How did they do it? In this paper, Netuitive shares some of their “how to” advice on:
- Common design mistakes to avoid
- Getting key Application Owners on board
- Delivering on performance promises by leveraging predictive analytics for IT
Lesson 1: “Why are we doing this again?”
When you embark on a potentially game-changing but difficult initiative, it’s important to remember why you’re doing it.
One global banking customer which Netuitive is working with puts their cloud computing initiative under the strategic umbrella of “Green IT” because it is easy to explain to industry analysts, the investing community, and the public. “We want to deliver the same services to our external and internal customers while consuming less energy, reducing our physical data center and carbon footprint, and saving money all at the same time.”
To achieve this end goal using a cloud computing model, IT management envisions the methodical development of a “unit” or “container” of computing power, based on virtualized resources available in private and eventually public clouds to minimize costs. Each “unit” would have an estimated operating cost as well as other performance parameters like power consumption and carbon footprint index. Once this computing “unit” is developed, the organization can meet with individual application owners to size up how many “units” their application needs – on average and at peak loads – to meet existing Service Level Agreements (SLAs). With a reliable range of monthly operational costs, this gives the application owners the flexibility to provision and configure how many “units” they want to fund for their application.
This bank did a thorough analysis and mirrored what other leading enterprises told us about how they categorized the expected benefits of their cloud computing initiative:
Avoiding the Bankrupt Private Cloud
- Provisioning and resource reallocation – new provisioning technologies for physical servers took the process from weeks to hours. In a virtual cloud, this can be reduced to minutes while adding the flexibility to dynamically “size” or provision virtual machines according to an infinite number of configurations to meet SLAs.
- Computing efficiency – server consolidation means fewer idle resources, and flexible and dynamic allocation means that applications should consume less power, on average – leading to better overall computing efficiency.
- Automated change control – compared to similar technologies for physical environments, virtual cloud environments deliver server configuration changes in minutes vs. hours. The same goes for configuration rollback, which is important since many incidents are caused by problematic configuration changes.
- Disaster recovery – the surest disaster recovery plans involve full “hot” backups of entire application infrastructures; however, these solutions are extremely expensive. Many companies believe that cloud based approaches to DR can deliver quick results at a much cheaper price tag.
- Peak load handling – provisioning applications for peak loads is extremely costly and inefficient. To prevent performance problems due to capacity, cloud infrastructures handle planned and unplanned peak loads by dynamically provisioning additional resources from private or public clouds on a temporary basis.
The lesson: it’s not just all about server consolidation.
Virtualization executives also note that, while server consolidation can save capital, it is often viewed as a tactical project within IT. The initiative becomes strategic when the vision is broadened to include running Tier 1, critical applications in your private cloud. This is where you need to remember the broad ranging benefits you will achieve for your enterprise to get you past the inevitable bumps along the way – like dealing with virtualization sprawl.
Lesson 2: The impact of virtual sprawl is worse than you think.
The ironic thing about virtual machine sprawl is that it can happen right under your nose.
The CTO of one large bank discovered the problem when his virtualization team was using Netuitive to optimize VM to host consolidation ratios. While looking for under-utilized physical assets, the team was surprised by a flood of alerts identifying idle or under-utilized virtual machines. The CTO states: “Be careful, because while you may set out to eliminate under-utilized physical assets, you may end up with virtual server sprawl. And it can cost just as much or more to monitor a virtual server as a physical one. Also, we found that we were too conservative on VM density to ensure performance, and we ended up over-provisioning just as we did before.”
Once they eliminated these over-provisioned VMs (which were often running idle), the bank was able to confidently increase their consolidation ratios from 5 to 15 VMs per virtualization host for Tier 2 and Tier 3 applications. The resulting additional millions of dollars saved in hardware and storage funded expansion of the virtualization project into a full-blown cloud computing initiative.
In addition to helping ensure maximum consolidation ratios, the VP at another bank reiterates why it’s so important to prevent VM sprawl. “Conventional wisdom is that managing a server running multiple VMs is similar to managing a single physical server. Not so: managing 20 VMs that share a server requires the same amount of work as 20 physical servers.”1 So eliminating VM sprawl not only minimizes hardware costs but significant operational costs as well.
The lesson: fixing virtual sprawl is crucial. Your infrastructure will be more efficient, you will be able to optimize consolidation ratios, and you will also save on systems management labor.
Check back tomorrow for Lessons 3 and 4.
- Predictive Analytics
- Application Performance Management
- Cloud Computing
- Management Team