Apply today for a FREE subscription to CIO Magazine!
Fri, Mar 28, 2008 20:12 EDT
|
Posted by: Venkat Devraj in Best Practices Topic: Infrastructure
Current Rating: |
Virtualization has grown in popularity as a means for organizations to reduce costs and increase efficiencies within the data center. Companies are adopting virtualization to consolidate and optimize their infrastructure, reduce the number of physical servers and associated real estate, power, and cooling costs, as well as reduce the complexity required to manage large numbers of servers and improve business continuity.
At first blush, virtualization’s inherent ability to commoditize data center assets such as servers and applications appears to reduce the need for automation. The hypervisor that turns your physical server into multiple virtual environments, seems to make time-consuming provisioning, disaster recovery and load balancing operations fast and predictable – often at the click of a button. Servers are managed from a central console, and the health of your virtual infrastructure is readily available from a single interface. Doesn’t this render data center automation obsolete in the virtualized world?
The fact is that the advent of virtualization has created a much larger demand for data center automation. There are many nuances to deploying virtualization and deploying virtualized environments actually creates more complexity, creating more room for error.
As the number of virtual environments increase, the corresponding need for caring, feeding and nurturing these environments also increases. For instance, compliance, security, configuration management and performance management activities increase manifold. If you thought patching your 1,000 physical servers posed a significant challenge, trying doing the same to your 5,000 virtual environments!
Also consider vendor application support. Application support on virtualized platforms is seldom defined by application vendors, and when it has been defined, the supported hypervisor could be from any of a variety of virtualization vendors. If you plan to push more of your applications to virtualized environments, you will need to run them on the supported hypervisor (think VMware, Microsoft Virtual Server, Citrix XenSource and IBM PowerVM).
Owning multiple flavors or versions of hypervisors requires additional staff with specific expertise to administer them and the lack of standardization across hypervisors has created several methodologies to perform seemingly identical operations. As the virtualized environment grows, you will find that heterogeneity and complexity grow in lockstep. Plan to invest upfront in additional staff and training, as well as to keep up with the eventual server sprawl across disparate virtual environments.
Automation software will play a large role in the virtualization explosion to help reduce the cost and complexity of managing virtual infrastructure. Data center automation solutions for virtualized environments can be quickly compared and sized against one another if you consider the following three criteria:
1. The automation software solves a problem specific to your environment very well and has demonstrable ROI.
2. The automation software serves all of your hypervisors, and can be extended to support additional hypervisors you may invest in down the road.
3. The automation software works for your existing physical infrastructure as well, and integrates with your existing systems to support other non-virtualization initiatives.
The third item is an interesting one to consider; return on existing assets means investing strategically to reap additional investment from your existing physical infrastructure - beyond the hypervisor. A good data center automation vendor will provide out-of-the-box automation to administer and provision both physical and virtual infrastructures as well as tie into critical initiatives such as ITIL-based configuration management, change management, incident management and problem management.
To sum it up, virtualization is a cost-effective method of achieving higher service level agreements while reducing your risk as your data center footprint continues to grow. However, expect heightened operational costs as well as a larger
I agree with the premise here… IT is progressively getting more complex every year. Virtualization is just another wave of complexity that is being added on top of already existing ones – mainframes, client server architecture, web services, SOA, etc…. with more coming in the future, and all of them interconnected in some shape or manner. Actually, after talking with many CIOs, it is rather obvious to them that it is humanly impossible to make sense of this complexity and properly manage SLAs of business applications and data centers with current tools and methods.
So yes, automation is an answer. However one can automate manual tasks and still just do things a bit faster than before.
What is needed is a new method of looking and consuming this massive amount of data and interconnectivity that drives complexity. What if a system can read through all the IT data and determine what’s behaving normally and what’s abnormally, and then correlate that behavior to the IT and business data and processes and provide the application owner with a forensic capability to take a fingerprint of the problem or situation that contains the building patterns of events that lead to the issues? Wow! Such a system would revolutionize the way IT is done today.
Fortunately, we’re starting to see these systems popping up more and more. Integrien is one of such revolutionary thinking companies that uses brute force math through probabilistic statistical cycle determination and correlation to solve complexity. So automation is good but it has to be coupled with a different way of thinking. Automating a deterministic process will lead to a slight increase in the ability to manage more. Automating IT using a non-deterministic probabilistic approach of determining normal and eliminating abnormal, coupled with the ability to massively correlate all IT data to all business data would lead to a ten fold increase of the ability to manage complexity and provide sane lives for IT professionals...