Infrastructure 2.0: Virtualization and Blades - Building for the Future

By Barb Goldworm
published: Thursday, May 29 2008

As the virtualization chair and keynote speaker this month at the Blade Systems Insight annual Conference in Tucson, I spent an interesting week talking with hundreds of IT executives (and a few vendors of course) about the state of both virtualization and blades today. Although the conference was primarily focused on blades, virtualization was a huge common thread that ran through all the keynotes and much of the conference. Here are some of the highlights of the conference, including thoughts from my virtualization keynote presentation, other keynotes, and conversations during the week.  

Green Platforms for Consolidation

Server consolidation has been a recent top priority project for large percentage of IT organizations. These consolidation efforts have been driven by the need for cost reduction, green initiatives, and in some cases either an attempt to avoid building a new datacenter, or simply running out of power.  Power, cooling, and green issues overall were key issues at the conference and for most of the IT execs that I talked with all week.

By consolidating servers and reducing the number of servers required to do the same workload, IT can reduce the space required for server farms, and reduce the power and cooling required to drive those servers. With typical server utilization running between 2 and 10% (before virtualization), consolidating through virtualization brings extremely compelling business benefits, with consolidation ratios typically running an average of 10 to 1 (consolidating 10 existing server workloads as 10 virtual servers onto one physical server).  Driving utilization up to 60 or 80% drives up the ROI (getting more value for the money you spend), and buying, managing, powering/cooling fewer servers brings TCO benefits in both initial expenditures (CAPEX) and ongoing expenses (OPEX). These benefits have been one of the most significant factors in driving server virtualization adoption rates up to the 80-90% they have reached in large and very large enterprises (although only 10% of all servers are estimated to be running virtual today).  Server virtualization solutions have matured, and offer stable platforms to allow organizations to continue rolling out virtualization in production across most of their remaining infrastructure.

These same space, power and cooling issues have also been a factor in driving server blade adoption, although there has been much confusion about power and cooling related to blades.  Certainly the high density of blades has been well understood, and IT shops with space issues were early adopters of blades, allowing them to pack more compute power per square footprint, and often avoid datacenter buildouts while continuing to grow their server capacity.  But power and cooling issues also have been a barrier to blades, both in reality and through misconceptions. In reality, first generation blade systems did have power and cooling issues, which caused many early lookers to stay away. Newer generations have corrected these early problems, and in fact, today's blades are more efficient in terms of power and cooling than their rack server counterparts. The reality now is that if you compare the power and cooling requirements of a single blade server to its equivalent rack server, it is will require less power and less cooling, and run more efficiently.  The challenge comes in when you then add the density factor, and pack so many more servers into the same physical footprint. So if you're looking for overall efficiency and power and cooling reduction, blades are a great answer. If your issue is power and cooling maximums within a certain footprint in part of the datacenter, blades will exacerbate rather than alleviate your problem.

Changing landscape

In the earlier days of blades and virtualization, organizations looking to solve their space, power and cooling challenges often looked at blades and virtualization as an "either/or" answer. Either they could virtualize, or they could move to blades to reduce space, power and cooling.  One reason for this was that early blades were really too lightweight for virtualization. In other words, they lacked the horsepower - CPU, memory and I/O connectivity - to be a good platform for virtualization. Again, as blades have matured, they now have the same options as rack servers, in terms of increased number or cores, increased memory, more NICs and more HBAs, to the point where in most cases, anything you can get on  a rack server, you can get on a blade server (depending on the vendor).  As a result, there has been a shift from the "either/or" to an "and", where blades are now becoming a platform of choice for virtualization. In fact, organizations that are virtualizing are now twice as likely to be implementing blades as those in the general market. 

Every keynote and user presentation I heard at the conference talked about virtualizing on blades. The opening user keynote talked about how the NFL builds the movable IT infrastructure for "Super Bowl City" via virtual servers on blades.  (The keynote of course opened with inspiring NFL footage - making the audience want to cheer for football, blades and virtualization.) Other users talked about the significant savings they've seen from virtualizing on blades, in space, power, cooling, and management, as well as the improvements they've seen in IT operational efficiency and rapid provisioning.

Another trend common across both technologies is the move from servers to desktops.  As organizations gain experience and have successes on the server side, many organizations are now re-evaluating their desktop strategy, and looking for ways to leverage these technologies to address the management, security, support and sometimes environmental issues involved.  Workstation and PC blades, and desktop and application virtualization offer a variety of options to help address the varied use cases for desktop users.  This shift to solving desktop challenges is just beginning, but will become an even bigger issue than servers, as it emerges, given the difficulties involved, the huge potential benefits, and the sheer number of desktops today and going forward.

Infrastructure 2.0 - a Paradigm Shift

The business case for both virtualization and blades around consolidation is clearly compelling, and most users who have implemented to date have done so based on consolidation.  For those shops that have the luxury of looking farther into the future, I believe these technologies hold much more promise.  These technologies allow IT  to eliminate the dependence on individual hardware components. They are the enablers for building a new type of IT infrastructure based on virtual resource pools which reside on modular, scalable, fungible hardware components to deliver on-demand, automated, rapidly provisioned, dynamic IT.    With stateless, fungible, remotely manageable servers, connecting to shared networked storage, and an overall virtual infrastructure including network, storage, server, desktop and application virtualization, IT can automate the infrastructure to do the work - provisioning new applications and services in months rather than minutes.  These are the infrastructure technologies of the future that can change the IT paradigm, and put IT back in the hero position - quickly delivering the services the business needs, rather than IT being the bottleneck of the business.

Blade Systems Insight Annual Conference ,  Focus on Virtualization