Brian Wilson of Quest Software presented best practices and pitfalls to avoid when deploying and managing a private cloud. Like many people, I wondered what expertise Quest had in the cloud - until he clarified that Quest recently acquired Surgient, which helps push them into the private cloud automation market.
Interestingly enough, you won't find the word 'virtualization' in the NIST cloud computing definition. That's because virtualization is a technology, not a characteristic of the cloud. According to Brian, there are five pitfalls and five best practices to consider when moving to a private cloud.
Pitfall #1: Confusing virtualized infrastructure with private cloud
Virtualization alone does not equal a private cloud
Focus on cloud benefits - self-service, self-management, pooled infrastructure, measured service or chargeback, on-demand access anywhere, rapid elasticity
Pitfall #2: Don't wait for the 'big bang'
Virtualize key apps, pool resources, virtualize more, pool more and repeat...
Avoid waiting for that magic metric
Evaluate processes - keep asking 'is there a better way to do this?'
Pitfall #3: Failing to understand self-service - different users are going to look at self-service differently, different groups and resources will define it differently within organization
Consider models with finite resources and defined timeframes (airlines, hotels)
Determine the importance of reservations (not requests) for your infrastructure
Determine the right balance of automation and control
Define appropriate policies, let it run, and evaluate success
The trouble with scheduling without guaranteed reservations - you must be able to deliver! The user experience must be complete. You have to guarantee reservations.
Pitfall #4: Understanding your users' definition of production
Understand what your user demands and expectations are - determine what redundancy, monitoring and support SLAs are necessary
Discuss your organization's definition of production vs. mission critical
Cloud technology today is ready for true production in most cases
Pitfall #5: Ignoring application lifecycle management
Create a plan (or even a team) up front to manage applications
Create reusable building blocks of images and applications
Find the sweet spot for unique number of images in your organization
Remember - content is your carrot!
Best Practice #1
Define your use case
Training - internal to staff or external to clients or other stakeholders
Sales evaluation and proof-of-concept
Best practice #2
Identify the 'right' applications for your cloud
Not all applications should be in the cloud
Characteristics to avoid: impossible or very hard to virtualized, hardware dependency, static over very long periods of time, specialized testing scenarios
Best practice #3
Know your users and plan their experience
Take the time to understand how they will (and could) use the cloud
Avoid confusion from too many or too few constituent groups
Understand that things WILL change - be flexible
Best practice #4
Prepare for compute platform and storage growth
It's almost too easy to grow
Determine scale points upfront to avoid future headaches - understand how you are going to manage your environment when you move to private cloud and how to scale
Be prepared to grow - quickly!
Best practice #5
Future-proof your cloud
Start with a defined project; plan for scale, heterogeneity, and change
Stay in touch with your vendor's roadmaps