Virtualization Technology News and Information
Article
RSS
VMTurbo: 2012 - The Birth of Unified Virtualization Management (UVM)

 

What do Virtualization and Cloud executives think about 2012? Find out in this VMblog.com series exclusive.

2012 - The Birth of Unified Virtualization Management (UVM)

Contributed Article by Yechiam Yemini, co-founder and chief scientific advisor, VMTurbo

A central challenge of IT is how to increase virtualization penetration rates from current 40% to fully-virtualized IT (80% and above).  According to the Gartner Group: "The use of x86-architecture virtual machines has been increasing at a torrid pace since their introduction in 2001; however, this growth will start to slow in 2012" (http://www.gartner.com/DisplayDocument?id=1573914). 

This slow down emerges as virtualization exhausts the pool of performance-insensitive applications, hitting the complexity barriers of virtualizing performance-sensitive mission-critical applications.  Leaving these mission-critical applications non-virtualized can result in costly, and difficult to manage, schism as virtualized and non-virtualized applications pursue different infrastructures and operations trajectories.

Unfortunately, current virtualization technologies are inadequate to provide the performance assurances, required by mission critical applications.  Virtualization is set to optimize efficient resource utilization through sharing. Such sharing comes at the cost of potential interference, disruptions and performance degradation. Performance insensitive applications can tolerate such occasional disruptions. In contrast, mission critical applications have low-tolerance to such disruption.  Unified Virtualization Management (UVM) technologies, considered below, promise to support systemic performance assurance of virtualized mission critical applications, and thus resolve the Full-Virtualization challenge.

What Will It Take to Virtualize Mission-Critical Applications?

Mission-critical applications often require tight guarantees of availability and performance.  Virtual infrastructures already support an impressive array of high-availability architectures. These pure software solutions often offer flexible availability features, superior to non-virtualized hardware solutions.  Accordingly, the key weakness of virtualization is in assuring applications performance.

Traditional non-virtualized architecture can assure applications performance through several mechanisms:

  1. Dedicating hardware resources to the applications statically
  2. Over-provisioning these resources to handle peak workloads;
  3. Exploiting specialized high-performance hardware (e.g., fabric, storage) to accelerate performance; and
  4. Empowering applications administrators to fully control these resources to optimize apps performance.

Virtual systems, in contrast,

  1. share resources dynamically among competing workloads leading to efficient resource utilization at the cost of mutual interference  and potential disruptions; a mission-critical application may see unacceptable performance degradation
  2. provide limited guarantees of resources allocated to an application; one may use Shares to guarantee the priority of an application (virtual machine)  in accessing shared resources; but such priority does not guarantee the actual resource capacity allocated; alternatively, one can use Reservations to guarantee the capacity allocated to an application; indeed,  one can reserve resources capacity to service the peak-demand of  application; however, such hypervisor-based dedicated over-provisioning obviates the primary value of virtualization in optimizing utilization; it begs the question, why use a hypervisor to dedicate resources to an application, rather than provide it with its own dedicated hardware?
  3. limits access to specialized hardware features through the hypervisor software abstraction; and
  4. dis-empower applications administrators from controlling resource allocations to applications

 

New technologies are required to extend virtualization to assure the performance of mission critical applications without compromising efficient resource utilization.  In particular, these technologies need to:
  1. Provide adaptive resource allocation to assure performance

    Adapt the resources allocated to a mission critical application to assure its performance targets, while optimizing resource utilization. (Contrast this with wasteful dedicated peak-provisioning, or with interference-prone static resource management using Reservations and Shares.)

  2. Empower collaborative management by both applications and virtualization administrators

    Integrate resources and performance management to support flexible harmonious division of responsibilities and authorities.

  3. Exploit emerging high-performance hardware to accelerate mission-critical applications

    Emerging hardware products (e.g., storage, fabric, applications delivery controllers, network switches...) provide specialized capabilities to enhance virtual infrastructures performance. The device vendors often support access to these capabilities by management systems.  Technologies are required to harness the power of these emerging infrastructure components, by integrating their management with hypervisor resource management and applications management features to best assure applications performance.

 

2012:  The Birth of Unified Virtualization Management (UVM)

UVM technologies can greatly simplify the virtualization of mission critical applications through three technologies:

  1. Adaptive Performance Assurance

    UVM systems monitor applications performance and adapt the resources allocated to them to assure performance while optimizing resource utilization. In contrast with dedicated-peak-provisioning, applications get right-sized resources, dynamically, as required to assure their performance metrics. In contrast with current virtualization systems, the capacity allocated to a virtual machine derives from performance assurance levels of its applications.

  2. Unified Entire Stack Management

    Current management functions and systems partition management responsibilities along strict stack boundaries.  Applications administration functions and tools are separated from infrastructures administration. UVM replaces this partitioned management with unified whole-stack management.  UVM tools support flexible collaborative management of all stack layers by increasing visibility of stack layers by all administrators and supporting coordinated control.

  3. Integrated Infrastructure Management

    Current management tools and functions are typically partitioned along element boundaries, often leading to fragmentation of processes and to barriers on incorporating new infrastructure components, or changing existing ones. UVM technologies will support integration of vendor-specific element management with hypervisor-specific management, creating unified management abstraction. These common abstractions can provide increased visibility of the infrastructures operations, well beyond current hypervisors boundaries. Furthermore, they can lay common grounds for streamlined unified whole-system management.

Conclusions

Expect UVM technologies and solutions to make in roads into virtualizing mission critical applications, starting 2012.  With time, UVM technologies will not only assure applications performance, but they will provide systemic whole-stack and whole-system management solutions well beyond the ad-hoc fragmentation and limitations of current systems.

###

About the Author

Yechiam Yemini is the co-founder, Chief Scientific Advisor at VMTurbo. He also is a Professor of computer science at Columbia University, where he has led research of networked systems. He has authored over 200 publications, holds15 patents and has lectured widely in these areas. Professor Yemini is a co-founder of three other companies: Comverse Technology; System Management Arts (SMARTS) acquired by EMC in 2005; and aROOTz, a start-up developing technologies for scalable video distribution over the Internet. He has served as a director and advisory board member of several high-tech companies and venture funds, and as a member of several government technology commissions and working groups.

Published Friday, November 11, 2011 5:30 AM by David Marshall
Filed under:
Comments
VMTurbo: 2012 – The Birth of Unified Virtualization Management … » Cat Back Exhaust Information - (Author's Link) - November 11, 2011 8:46 AM
VMTurbo: 2012 – The Birth of Unified Virtualization Management … | Web Hosting Geeks - Shared Web Hosting, VPS, Dedicated Servers, Virtualization and Cloud Computing - (Author's Link) - November 11, 2011 3:00 PM
VMblog.com - Virtualization Technology News and Information for Everyone - (Author's Link) - January 4, 2012 7:08 AM

I'd like to personally welcome each and every one of you to the start of 2012! As we begin what will certainly prove to be a fantastic new year, I wanted to make sure to thank all of the loyal member's and readers of VMblog.com. Once again, with the help

2012 – The Birth of Unified Virtualization Management (VMblog) | VmTurbo - (Author's Link) - June 1, 2012 12:05 PM
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
top25
Calendar
<November 2011>
SuMoTuWeThFrSa
303112345
6789101112
13141516171819
20212223242526
27282930123
45678910