Reducing Infrastructure Costs Through Virtualization
Leandro Sgallari
TI Sistemas
June 2009
Summary: Nowadays, when we think about projects, it is necessary that we have a precise budget and defined cost beforehand. This article identifies the virtualization model—at all possible levels—as the best fit today.
Contents
Introduction
The Original Model
Server Virtualization
Desktop Virtualization
Virtualization of the IT Department
Cloud Computing
Example of Architecture
Conclusion
Introduction
Nowadays, we live in uncertain times all around the world. When it comes to architecture and design, we must think a lot more in costs—unlike in other times, when we used to think first of the solution and then in costs. If we had a streamlined and defined return on investment (ROI), only then did the project get the green light. Today, if we think about projects, it is necessary to have a precise budget and defined cost first before we can start to think about the project.
Upon brief reflection of what IT architecture is, one finds that the model that the whole world favors (because of costs and the evolution of technologies) is the virtualization model. Ten years ago, it was all about decentralization—both in data centers and servers, and in communication and desktops. When we thought of an application, we always tried to have the layers of such an application as close to the client as possible; data centers, databases, and e-mail servers were distributed all over.
This situation was the result of being unable to face the huge cost of having hardware equipment of multiple large capabilities (such as an eight-processor server and lots and lots of RAM gigabytes) or having redundant point-to-point communication links with good bandwidth—the cost of which only large companies could consider including in their architecture. All of this encouraged a decentralized IT administration model that required specialists in the different platforms of each site.
The Original Model
Although this model worked for years, many things were not considered that today have rendered this model not as efficient as was initially thought.
Figure 1. Typical scheme
Taking as an example the analysis of a distributed application that was devised 10 years ago, there was an architecture in which it was important to have the data near the client, which led to the following scheme:
A database in the central office in which the information from the different sites, the database from each site, the application server from each site, and the local applications that were installed on each desktop were all consolidated.
A replication scheme among the databases was used for the distribution of information. This drove us to have database administration for each site, besides having on each site an infrastructure administrator who had thorough knowledge.
Initially, this had (as variables within the equation) high communication-link costs, large servers that represented a very high cost, and operative systems that were neither very solid nor rigid with regard to changes and also had little functionality—that is, they offered very few functions or roles within the operative system.
For many of the needs of the company, it was necessary to add software that could comply with that functionality. In addition, in order to carry out the tiniest of changes, it was necessary to set the server offline and have IT personnel who represented an average cost. The variable that was not really considered was the updating and maintenance of the whole structure, which at the time—due to the fact that technology did not evolve in the way that it does today—was not such an important aspect.
If we consider basic accounting principles (which I have learned during recent years), one should always see the IT personnel as an asset to the company, with both amortization time (which is the time that it takes to shape the person, according to the culture and needs of the company) and an updating cost (which is what must be invested to have a person trained in the different technologies as they evolve).
Over time, all of this changed; the variables in this equation also changed, and the updating and maintenance variable (which in many cases had not been taken into account) started to gain more and more importance.
This is the equation that we face nowadays:
Average to low communication-link costs (taking into account the virtual private network), large servers with many RAM gigabytes at average to low cost, operative systems thathad hundreds of embedded and flexible options and lots of functionality (many things already come solved and embedded in the operative system, so that in general it is not necessaryto set the server offline to make these changes), averageto high personnel costs, and average to high updating and maintenance costs.
Within the scheme that is encouraged today, many things must be considered; it is necessary, therefore, to have the whole scheme in mind—not just a part of it—to avoid making the same mistakes that we incurred in the past.
Nowadays, when uncertainty (crises, corporate mergers, acquisitions, and constant changes) is all around, it is vital to work toward an environment that would basically support constant dynamic changes. More than ever, it is necessary to think about platform and application updating, growth, and corporate and budget contractions. This, of course, will highly influence the model that is to be chosen, and that model (taking into account the aforementioned equation) should be based mainly on the updating and maintenance variable.
When we consider all of the preceding, we will see that the model that best fits is the virtualization model applied to all of the possible levels, where all of the equation variables are considered in order to determine feasibility and total cost of ownership (TCO).
There will be infinite virtualization scenarios—from choosing cloud computing in specific services and virtualizing (or outsourcing) the whole or part of the IT department to using virtualization for servers, applications, or desktops.
Server Virtualization
Today, there are many important players and technologies that have been widely tested, such as Microsoft Hyper-V and VMWare. The hardware costs have gone considerably down: If we were to compare four- to eight-processor equipment of the past to one today, it would result in an important cost margin that would be an improvement, and it would be necessary to add the progress that has been made in technologies and redundancy within equipment, board, hot-plug memories, and so on.
Generally speaking, almost all the components of the server can be changed without having to take the server offline. The same thing applies to operative systems. This means that we can do away with the theory that was used in the past, according to which we used to divide into different hardware pieces the different business applications. In addition to this, the advantages of tolerance to failure in the virtualization schemes that are used today make it possible to take a physical server offline without affecting the virtual server that is running in that physical server. This, of course, means that from a simple technical point of view, there is already a huge advantage in the use of server virtualization.
From an architectural point of view, this allows us to respond to organizational changes quickly. Above all (and this is something to consider nowadays), this also enables us to achieve substantial savings at the time of shaping a data center. There will also be savings in the electrical bill, refrigeration costs, physical space, and hardware. Moreover, deployment and disaster recovery will be much simpler.
All of this will result in lower maintenance costs—whether we have our own IT department in 100 percent administration of this platform or a virtualized environment of the IT department (later, we will develop the IT department cost, as well as the options and the reasons behind virtualizing it). This all leads to there not being almost any scenario in which virtualization is not applied and which will not result in a much lower TCO.
Desktop Virtualization
In this regard, it is also possible to find very well-developed and well-tested technologies, such as Microsoft Terminal Server, Citrix, and so on. This kind of virtualization was previously thought about for remote points or links that had a relatively small bandwidth. Nowadays, it is used as a method to reduce desktop-administration expenses, because (thanks to this technology) it is possible technically to have tolerance to failures, add it to the server-virtualization scheme, and create a pool of servers. There is substantial reduction of the desktop-maintenance cost and the cost of desktops themselves, because with equipment that has smaller hardware, it is possible to run any kind of application and still have centralized control and deployment of applications and security policies.
Virtualization of the IT Department
In the past, there was a relatively low or not-so-significant IT department cost, compared to the one nowadays. Today, there is a high IT department cost, and it is necessary to consider the following variables: training in new technologies, training in the company environment, the cost of personnel search, and the time during which the search takes place. All of this leads to the IT department not being able to respond with the speed that the company needs. In addition, we currently experience a high labor turnover, which means that many times this process has to start again—which, of course, drives the cost upwards.
If we consider all of these factors, especially the costs and the time that the search involves, we will see that having a virtualized IT department results in a lower TCO and in every possible advantage. With virtualization, all of these IT-department problems are moved to an external company that is exclusively devoted to IT, particularly as it refers to specialists in technology or in specific technologies. This would mean that there is no point in having a specialist as part of the internal IT department.
Cloud Computing
Currently, there an infinite number of services are available on the Web, from e-mail services (as has been the case for a very long time) to CRM, ERP, Document Managers, and other services. This solution naturally offers a world of advantages: It is unnecessary to have a specialist in this technology within our IT department, it is equally unnecessary to maintain that technology from either the hardware or the software point of view, and keeping security copies of the information is no longer required. Depending on the kind of hired service and service-level agreement (SLA), it will be possible to have a redundant and always-online service. In some cases, the cost for this kind of service can be high—depending on both the number of users within our organization who require this service and the characteristics of the service—and is worth considering.
Example of Architecture
(Example based on a company with 500 employees.)
Figure 2. Current architecture
As an example, we will use a virtualization architecture that uses such Microsoft technologies as Hyper-V and Terminal Server. Number of servers: 15.
Typical structure of IT department: One manager, two IT administrators, one database administrator (DBA), and five Help Desk employees.
Based on everything that was explained previously, we will take the best of each virtualization technology to carry out a cost reduction.
Server Virtualization
It will be possible to reduce approximately 15 noncritical servers into 4 physical servers that will be able to support these 15 virtual servers. It will be necessary to carry out a load analysis and distribute the servers and business applications correctly. Nowadays, given the kind of roles of a typical company server, there are not many servers that have a high processing consumption; thus, it will be necessary to isolate these kinds of servers, so that a resource competition conflict is not generated. It will also be necessary to include (if we do not already have it) an external storage in which our virtualization scheme will be stored (so that it is possible to work on it in a cluster) and which will have tolerance to failure of all of the virtual equipment. All of this will be possible by using Microsoft Windows Server 2008 64 bits and Hyper-V System Center Virtual Machine Manager to carry out the P2V conversions.
Cost Reduction
Within this example, it will be possible to reduce approximately 70 percent of the energy consumption, as a result of less consumption on the part of the servers. In addition, there will also be a reduction of approximately 70 percent in the refrigeration consumption, as a result of the use of storage.
The licensing cost will also decrease (when we use Microsoft licensing) very substantially. The Microsoft licensing scheme is based on Table 1. In the table, we can see that by using Windows Server 2008 data-center server licensing, it will be possible to obtain a reduction in licensing from 15 servers (which will be able to use different versions of Windows Server 2008, depending on the processor and RAM needs) to only 4 (with data-center or enterprise licenses). Depending on the versions of Microsoft Windows that are used, in the least favorable scenario, we will achieve a reduction in cost of 50 percent.
| Version of Windows Server 2008 host | Covered virtual servers |
| Standard | 1 |
| Enterprise | 4 |
| Data center | Unlimited |
Table 1. Microsoft Licensing Scheme
Desktop Virtualization
Depending on the memory consumption of the applications, it will be possible to implement approximately five virtual servers for Terminal Server— typically, in five physical servers to cover 500 work positions. The main advantage of having virtualized servers is that this will automatically commute to any other, in case of a failure in any physical equipment.
In this way, we will be able to have a desktop with fewer resources, and it will be possible to update the applications more rapidly, as with deployment, management of printers, and any other desktop problem. In turn, this will also enable us to make the desktop of the user available to remote or external users.
Cost Reduction
If we consider, on the one hand, the cost of updating 500 desktops as a result of the installation of some business application and, on the other hand, the purchase of five 32-GB RAM servers and two Quad Core processors each, we will obtain a cost reduction of approximately 90 percent.
Virtualization of the IT Department
It is first necessary to analyze the critical and noncritical applications; it is important also to analyze the IT labor market in the country in which it is applied. Generally speaking, the advice that is given is to virtualize whatever is difficult to get in the market and to have partial virtualization of the IT department. For this example—and considering the current work market—we will opt to virtualize (for example) only one IT administrator and one DBA; the Help Desk, one IT Administrator, and the IT manager will continue to be physical. By having an SLA with external suppliers and a framework work contract, it will be possible to increase rapidly the IT department or change swiftly the scheme without a great increase in initial costs. It will also be possible to decrease training costs, hiring costs, and so on.
Figure 3. Virtual-server scenario
Cost Reduction
If we consider the TCO, hiring costs, training costs, and salaries, we will obtain a cost reduction of approximately 55 percent.
Cloud Computing
Let us take an application that will not be worth having internally, because of the size of the company. For this example, we will use a CRM. Ten CRM licenses will be hired online. In this way, no costs will be associated to the initial licensing, administration training, CRM server deployment, disaster-recovery policies, or anything that pertains to administration of the CRM.
Cost Reduction
Based on online services, there will be a cost reduction of approximately 80 percent—based on 10 licenses, and considering the initial cost of having a server, trained personnel, backup policies, and so on.
Conclusion
The virtualization scenario makes it possible to make structural changes in the IT department with the speed that the market actually needs.
We can have strong cost reduction, because with a physical structure, we often do not use all of the resources—hardware, software, employees, and so on—at 100 percent. On the other hand, with virtualization we have the opposite case: We use and push the resource utilization as far as possible, and then we add more resources to virtualize.
In our example, we can see the individual cost reduction; if we look at it globally, however, the cost reduction is more significant.
About the author
Leandro Sgallari (LeandroS@tisistemas.com.ar) is owner of and consultant at TI Sistemas in Argentina and has over 12 years of experience in IT infrastructure. He is project leader and infrastructure architect for IT projects. Leandro is a Microsoft Certified Trainer and teaches Microsoft Windows 2000/2003 official courses in Microsoft Certified Partner(s) for Learning Solutions (MCPLS). He is a Microsoft Certified Systems Administrator (MCSA) and Microsoft Certified Systems Engineer (MCSE) in Windows 2000/2003, as well as an early achiever in Windows 2000/2003, Microsoft Certified Trainer (MCT), Microsoft Certified Database Administrator (MCDBA), and Solutions Architect MVP. Visit Leandro’s profile.
Follow up on this topic