The ROI for Server Virtualization with Business Critical Applications

Server virtualization has been pretty much of a no-brainer for many IT departments as they were able to dramatically reduce the number of physical servers under management. This has been possible since IT has been able to virtualize the applications that IT owns and supports without IT having to consult or sell teams that own business critical applications. The ROI’s from virtualization have been driven largely by the savings associated with server consolidation. However, as these same IT organizations start to virtualized business critical applications different dynamics and assumptions come into play.

This post compares the ROI’s on two server virtualization projects – the first assumes tactical applications with significant server consolidation (actually core consolidation), and the second assumes business critical applications with no server consolidation (again actually no consolidation in the number of cores per workload). Using VMware’s ROI Calculator shows us how the ROI’s varied for these two cases.

The Tactical Applications Case (With Server Consolidation Savings)

For the sake of this analysis we assumed the following:

  1. 100 workloads were to be virtualized, each of which is running on its own physical server prior to virtualization.
  2. The servers in the physical environment are at the end of their 5 year useful life. So the choice is to buy 100 new physical servers, or to virtualize and buy fewer servers
  3. The existing physical servers are in a dual processor dual core configuration
  4. There are 100 servers X 2 CPU’s per server X 2 Cores per CPU or 400 cores in the configuration for the 100 workloads (4 cores per workload)
  5. The new servers for the virtual environment will be in a dual processor quad core configuration (8 cores per server as opposed to 4 cores per server for the servers to be retired)
  6. Once virtualized, each workload will get one core resulting in an 8:1 server consolidation ratio (4:1 due to core consolidation and 2:1 due to 2x the cores per server)
  7. There is no growth in workloads (just for the sake of making the analysis simple)

The Server Configuration comparison for the physical vs the virtual case is shown below:

The savings from server consolidation are immediate and compelling. In year one of the project instead of spending $408,925 to buy 100 new servers for a physical environment, one only has to buy 13 servers to support the 101 workloads (one workload for Virtual Center) for $79,740. This saves $329,185 in hardware purchase costs in the first year with no more hardware purchases required for the next four years as shown below.

The savings in hardware drive the entire ROI of the project. After you go though the entire calculator and work through Power, Cooling, Networking, Rack Space, Storage, add in HA, DR, FT, and the rest of the vSphere Enterprise Plus features and their licensing, you end up spending $635,138 and saving $1,104,189  with a net positive ROI of 74% in the first three years as shown in the ROI summary below:

The Business Critical Applications Case (No  Core Consolidation)

For the sake of this case we assuming the following:

  1. The applications that are being virtualized are business critical tier 1 applications
  2. The physical environment is sized appropriately to be able to deal with periodic and seasonal spikes in demands upon the application
  3. The number of cores per workload will therefore not change once the application gets virtualized. It will remain at 4 cores per workload
  4. As above, the existing servers are at the end of their useful life and new servers get purchased
  5. Since the new servers have twice as many cores per server, only 51 new servers with vSphere are needed as opposed to 100 new physical servers
  6. The 8 cores per server lead to a 2:1 consolidation ratio for servers, but no consolidation for cores per workload (still 4:1 as opposed to 1:1 in the tactical applications case)

The Server Configuration comparison for the physical vs the virtual case is shown below:

Since old 4 core servers are being replaced with new 8 core servers a smaller number of total servers is needed in order to support the new environment. As opposed to spending $408,925 to buy 100 new servers, only 51 are needed at a total cost of $312,828 which saves $96,097 in server purchase costs as shown below.

However, the smaller savings in hardware create an ROI problem for the project. This time after you go though the entire calculator and working through Power, Cooling, Networking, Rack Space, Storage, add in HA, DR, FT, and the rest of the vSphere Enterprise Plus features and their licensing, you end up spending $1,674,838 and saving $803,750 with a net  ROI of (52)% in the first three years (the project looses $871,088) as shown in the ROI summary below:

Summary

This analysis took a very simplified approach to illustrate what happens to the ROI for a server virtualization project when it turns out the the team that sized the physical servers appropriately purchased the right numbers of cores per workload. When this is the case the only basis for server consolidation is the replacement of old servers with newer servers that have more cores per server resulting in a need for the same number of cores, but fewer physical servers. To understand the impact of this, it is best to illustrate the two ROI summaries side by side.

Tactical Applications Case Business Critical Applications Case

If this analysis is at all representative of what will be the case as business critical and performance critical applications get virtualized this raises some significant challenges for VMware and IT organizations who seek to justify these projects:

  1. Relying upon just server consolidation (or more specifically upon the number of cores per workload as having been over-provisioned) when virtualizing a business critical application will not result in an acceptable ROI if the physical servers and the number of cores per workload were appropriately sized in the first place.
  2. This analysis did not include the likelihood that there will be significant additional expenses associated with virtualizing a business critical application that were not present when virtualizing a tactical application. These include the need for more robust security management, configuration management, infrastructure performance management, role based provisioning and administration, and finally virtualization aware applications performance management. This entire layer of management software and the cost of procuring it and owning it was not included in this analysis.
  3. The way in which virtualization is “sold” therefore needs to significantly evolve. If up front savings in capital expenditures are not going to drive the economics of the project, then something else will have to take its place. This is likely to be something in the area of technical agility for the IT Operations group and business agility for the applications teams and business constituents, but a great deal of work needs to be done to formalize these benefits into something that vendors and customers can hang their hats on.

Enterprises struggling with these issues are encouraged to look at the question of how to manage dynamic (virtualized) systems and applications in a holistic and consistent manner. For many business critical applications, there is a silo of processes and tools that have been built up around each application – leading to tremendous inefficiencies in processes and tool use across large numbers of applications. Virtualization creates an opportunity to manage a range of different business critical applications in a more consistent, scalable and cost effective manner. It is likely that in order to achieve the operational savings required to justify virtualizing these tier 1 applications that the silos of processes and management tools that have been built up around these applications will have to be reinvented and reimplemented around a new set of virtualization based processes and tools. There are many vendors in the VMware ecosystem who offer excellent solutions in this area, and some very good ones are profiled in our Solutions Showcase.

Bernd Harzog (310 Posts)

Bernd Harzog is the Analyst at The Virtualization Practice for Performance and Capacity Management and IT as a Service (Private Cloud). Bernd is also the CEO and founder of APM Experts a company that provides strategic marketing services to vendors in the virtualization performance management, and application performance management markets. Prior to these two companies, Bernd was the CEO of RTO Software, the VP Products at Netuitive, a General Manager at Xcellenet, and Research Director for Systems Software at Gartner Group. Bernd has an MBA in Marketing from the University of Chicago.

Connect with Bernd Harzog:

Tags: , ,

4 Responses to The ROI for Server Virtualization with Business Critical Applications

  1. July 14, 2010 at 10:24 AM

    Bernd,

    I agree with the virtualized tier 1 application justifications and process reengineering you layout in your Summary section. I do have a couple of issues with the simplified ROI scenario you present. The last time I looked at the VMware VMmark scores, for instance, was around a year ago not long after the Nehalem came out. Even then, Dell showed a VMmark score of 24 which I believe translates into a consolidation of around 24 (average) VMs per 2-CPU server. With today’s more advanced Intel chips, we see an overall average consolidation of around 15 – 20 VMs per 2-CPU host including Tier 1 applications.

    Also, you take the approach that an organization is going to virtualize its tier 1 applications as a silo prodject. The reality is that the tier 1 apps will layer on top of the existing virtual infrastructure investment including hosts, storage, network, licensing and staff expertise. Particularly when deploying powerful technologies designed for virtual hosting such as the Cisco UCS, an organization can leverage its existing infrastructure to inexpensively add additional VMs – including Tier 1 applications.

  2. Bharzog
    July 14, 2010 at 10:32 AM

    Steve,

    I agree with the issues you raised. I tried stipulate in the post that this was a simplified analysis. One of the simplifications is indeed the look at just one Tier 1 application, not N in combination, nor N on top of an existing environment. I also agree there is more consolidation possible due to Moore’s law than I laid out, and agree that technology convergence (UCS) allows additional consolidation to occur.

    But my fundamental point is that the ROI for tactical applications is “back of the envelope” compelling, and the ROI for Tier 1 applications requires a lot more thought, work, and may in fact be driven more by OPEX than CAPEX.

    Bottom line is that I think that this is where the Virtualization Management tools will come into play, and where virtualization ceases to become a hypervisor play, but becomes a play about how you manage a dynamic data center across the sets of people who need to be involved.

Leave a Reply

Your email address will not be published. Required fields are marked *

Please Share

Featured Solutions