By Dan Kusnetzky
published: Wednesday, March 19 2008
By Dan Kusnetzky
Overview
Virtualization has tremendous advantages over physical environments, but in the move to virtual, the organization often lost some things it took for granted. One of those things is clarity of what is going on, where, and with whom.
What's WHAT?
The midrange systems from Sun or Hewlett-Packard are over in the corner running manufacturing or billing applications, the mainframe that's sitting in the center of the datacenter is a whiz at processing thousands upon thousands of transactions, and a rack of industry standard (x86) machines are supporting Web-based applications and the organization's collaborative systems.
As organizations began moving functions from physical systems to virtual systems they ran straight into a new problem. Virtual machines are ephemeral. They can be generated, provisioned with the appropriate software and put into production very rapidly. They can be halted and deleted when they no longer are needed. No labels define their location or presence.
IT executives are increasingly facing the fact that there really is no good way of telling what virtualized applications are running, where they are located at any given moment, which business unit owns them, when they were created, when they should expire, which physical resources they are using and whole host of other questions. It is no longer easy to determine either what physical systems are doing or, more importantly, not doing.
What's Where?
If an organization needs to maintain an audit trail of where the computing was done, where the data is and other important data points, they increasingly face
a very difficult challenge. Complying with some regulations may be impossible unless there is a structured, well-defined way to track everything. Organizations simply must have a trail of their motion, a history, or a chain of custody as they transfer from place to place and from test beds to actual production.
What's old?
Since virtual resources may be created easily, live only as long as needed and then be easily retired, the virtual machines' born-on dates and when they are to be terminated is now important information that IT administrators must track. Since this is a new function, traditional products and processes are not likely to be of much help. Will administrators fall back on privately maintained spreadsheets? Will this approach pass muster when an auditor comes? How easily will that scale when there are 5 or even 20 administrators in different facilities able to create virtual machines?
It is clear that a better way to manage this type of information is needed. Otherwise, organizations will find that they've created a virtual machine sprawl that will add additional challenges rather than reducing or removing them.
The Golden Template Concept
In order to create some order out of this potential chaos, many organizations have begun creating golden template virtual machine files. Each of these templates acts as the foundation for many other virtual machines. Golden templates are created using all of the necessary infrastructure software needed by applications. While this approach has some benefits, managing these elusive golden templates can also be a challenge.
Many organizations opt to use clones, rather than templates, to generate virtual machines, While the solution is different, the same questions remain: those of identification and virtual lineage of each virtual machine in the environment.
Organizations need to know the following information about each and every golden template or clone, every update to that template or clone, and, in turn, all of its children:
- Creation date
- Operating system type
- Operating system patch level
- Type of data management software
- Data management software version
- Data management software patch level
- Type of application framework software
- Application framework version
- Application framework patch level
- Applications
- Known incompatibilities
- Planned deletion date
- What children have been created based upon each template?
- Who are the sibling clones in the environment?
- Are there grandchildren as well?
Why traditional approaches don't work
For example, one simply doesn't install a physical system one morning and retire it that afternoon. Virtual systems are often created, used and then destroyed in a day's time.
Furthermore, it is unlikely that a physical system would be picked up and moved from one side of the datacenter to another several times a day. In a virtual world, this is quite possible. There are many vendors offering orchestration tools that can move virtual systems from one physical machine to another or change it from being a virtual resource to a physical resource based upon polices or to achieve service level objectives.
Not only is the fundamental data model of traditional tools flawed in these ways, the architecture is equally inappropriate. Fat agents in each virtual machine could be the end of any performance gains made by virtualization. And log files are often not sufficient for generating real-time insight into the environment. The challenge of monitoring virtual environments has many new dimensions of complexity.
It is not clear who would be responsible for virtual resources, as well. Would the system administrators who operate physical systems be also be responsible for the virtual resources that machine might be supporting? Would that job fall to the person in one of the business units who created and is using that virtual system? In the end, without help, administrators really have no reliable way to know what's running, where its running, and what compatibility issues it may generate.
Summary
As organizations increasingly turn to virtual machine technology for both client and server environments, it would be wise to consider the challenges this technology imposes as well as the benefits they may bring.
The Kusnetzky Group recommends that organizations develop a sound, well planned set of processes and procedures to manage these resources. It would also be wise to seek out tools that can automate this process. Suppliers such as Fortisphere are offering such tools. It would be wise of IT decision-makers to become familiar with the techology available in this category.
Daniel Kusnetzky has over 30 years of industry experience. He is responsible for research and analysis on open source software, virtualization software and system software. He examines emerging technology trends, vendor strategies, research and development issues and end-user integration requirements. In the past he was executive vice president for Open-Xchange, Inc., and Program Vice President of System Software Research for International Data Corporation.
|
