Are you getting the most out of Virtualization?

virtualization men.jpg

 

Guest post by

Sonu Sudhakaran,

Software Analyst, Virtualization and Cloud Solutions

 

This is my first blog in this Virtualization series, and am beginning with the challenges that are faced by Virtualization admins (this is hoping that admins have architected your IT) in delivering an infrastructure that meets the expectations of application administrators. After all, your real need is to keep your application admins happy all the time because they are responsible for offering the service to end users.

 

I grew up during the time when mainframes and super computers were evolving.  Applications were designed for scale and performance under a share-nothing model. That means, applications were guaranteed to get resources offered by server as a whole. In those days, administrators were involved in defining the application-to-resource ratio and were challenged with few concerns (like how many processes can run on a server with given capacity). But once the tuning or benchmarking is done, from that time the performance and response time of the application stays more or less constant on the server.

 

virtualization.pngBut that was the past…

 

The virtualization technologies (vSphere from VMWare, Hyper-V from Microsoft and KVM from Red Hat) have begun transforming how modern age data centers are designed or architected.  Administrators (System and Virtualization) now have the challenge of adopting this new style of IT and meeting the expectations of application administrators who have become more conscious and stringent with their requirements on performance, response time and scale. This is a shift from a “Shared Nothing” to a “Shared Everything” model of resource allocation and usage. The goal of  maintaining the same level of performance and response time for the application is now posing as a big challenge and often puts the virtualization admin red faced when App admins complain of not getting the assured resources when their applications need them the most.

 

To design a new age data center under the “Shared resource model”, some specialists use tools that allow “Physical to Virtual” migrations possible with just a few mouse clicks. Such tools usually work when you consider the static recommendations that virtualization vendors offer in their benchmarking guide.  For example, how many Virtual CPU resources to allocate per physical core and then offer a plan to migrate the server workload into one or many virtual machine(s) on the underlying hypervisor technology.  In many cases, application specification may also govern the spec of the virtual machine it will be hosted on. But in whole, the demand or resource usage comes least into consideration. And this results in “Resource Sprawl” (described later) and limits the capability of your IT.

 

To transform your datacenter and prepare for the next generation and realize the benefits from Virtualization technology, one needs to change your mindset. You need to consider “Resource Demand” and base your decisions of allocation of resources to virtual machine(s) on factors of resource demand than some static written specifications. In demand-based allocation, you consider peak usage of a resource by the workload running inside the virtual machine and then make the resource allocation to the virtual machine. This results in less resource waste and helps in achieving better consolidation ratio (Host to VM ratio or VM density/Host). Also results in a good balance of allocation and usage in your datacenter.

 

The following are few reasons that prevent enterprises from taking full advantage of virtualization technology.  Ignoring these facts may result in very little or no measurable gains in CAPEX that most virtualization vendors claim.

 

a)      Avoid Resource Sprawl (Perform more “Right Sizing”)

 

“Right sizing” is all about allocating virtualized resources (like CPU, Memory and Disk) and considering the usage of the workload running within the Virtual machine.

This is a very important concept to consider for the following reasons:

-          Flexing over-sized virtual machines (in terms of CPU and Memory) could help you solve some inherent performance issues the virtual machine is facing due to Hypervisor overheads (for example, Idle CPU(s) may result in bad CPU performance of the VM in case of VMware).

-          Identifying over-sized virtual machines and flexing their configuration as per their demand helps reclaim some otherwise wasted resources.

 

b)      Plan for Over-Commitment

 

One of the key benefits of using virtualization is the ability to treat CPU, Memory and Disk resources as commodity. Most vendors support over-commitment of these resources to a great extent. So make use of it! And plan for over-commitment to increase the benefits of Virtualization.

 

 

 

c)       Trend of Resource Usage and detect bottlenecks in your virtual environment

 

While a virtualization admin is excited about over-commitment and the flexibility of changing configurations without a big deal of effort (hot add CPU and Memory), special attention should be given to the fact that management of these resources is left to the underlying Hypervisor.

And you don’t want to over-commit resources so much that your IT doesn’t tolerate sudden increases in the usage of resources.

Analyzing the resource usage trend of each virtual machine in your IT and matching it against the available capacity plays a key role in taking out risks of over-commitment.

 

d)      Watchful Inventory management for elements in your datacenter (VM Sprawl, Storage Sprawl).

 

Virtualization vendors are making the job of Creating/Cloning/Moving of Virtual entities (to be specific VMs) very easy  Many orchestration tools (Puppet-Chef, HP Cloud Service Automation) have used the automation interfaces provided by these vendors to automate the creation of VMs upon the subscription request from their users.

And I have seen, users loosing track of their VM counts and virtualization admins have big trouble in figuring out the VM(s) causing sprawl. It is very important to eliminate:

 

-          Virtual machine sprawl – Limits the capability of Virtualization!

-          Virtual machine snapshot sprawl – Makes your storage management a nuisance!

It adds to the operational cost at the end of the day and limits your further allocation!

 

 

e)      Stay updated with upcoming capacity requirements for your datacenter.

 

One answer that every admin seeks is “How many more virtual machines can my data center accommodate?”

The expected answer is simple, but the factors that should be evaluated to answer this simple question are many. But while you plan your environment, you want to know what your limit is. After all, you are limited by the capacity of physical devices in your environment.

 

Now, it looks like adopting virtualization and using it effectively needs a lot of analysis and thorough monitoring of the current state of the data center.  Imagine a tool that gives you assistance in utilizing your virtual infrastructure to the best of its capacity.

 

HP Performance Viewer (vPV) helps in mastering the concepts described above which you would like to consider while using Virtualization in your IT. To learn more about this great tool, please visit the vPV Home Page.

 

You can also download HP vPV here to experience it for yourself.

Leave a Comment

We encourage you to share your comments on this post. Comments are moderated and will be reviewed
and posted as promptly as possible during regular business hours

To ensure your comment is published, be sure to follow the Community Guidelines.

Be sure to enter a unique name. You can't reuse a name that's already in use.
Be sure to enter a unique email address. You can't reuse an email address that's already in use.
Type the characters you see in the picture above.Type the words you hear.
Search
Showing results for 
Search instead for 
Do you mean 
About the Author
This account is for guest bloggers. The blog post will identify the blogger.


Follow Us
The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation