Corporate Computing – Alive and Well

by Sally Tomasevic, MBA

Disclaimer and Preface. First, Sally Tomasevic, the author of this article, is my wife (I am Marcel Gagné, the owner of this site). Second, Sally wrote this article as part of her MBA work, but with the constant news talk about cloud computing, both hype and reality, including the inevitable quotes from Nicholas Carr, I found that Sally's work was as relevant and important now as when she wrote it, nearly three years ago. With her permission, I publish it here. -- Marcel Gagné

Introduction

Are we now, as Nicholas Carr speculates in his provocatively titled paper, "The End of Corporate Computing", on the eve of an IT revolution? While Carr claims that corporations can share utility hardware and software and champions the cost effectiveness of utility IT computing in the future, he completely fails to consider necessary elements of computing that cannot be provided for within this model.

Carr's argument centres around similarities he believes exist between the history of electrification and the IT industry. In the early 20th century, private generators were housed and maintained at corporations and over time, while utilities and associated infrastructure were built, corporations were supplied electricity by utility providers. As Carr sees it, the IT industry is transforming from a similar asset based model in which IT components reside within the walls of corporations, to a utility based model wherein all IT services can be provided by large, centralized, utility providers. Carr believes that the technological advances in IT, specifically virtualization, grid computing, and web services, are sufficient to pave the way to a utility model. In addition, he views on site corporate computing infrastructure as redundant, postulating that within the utility model, thin clients will replace PCs on corporate desktops.

Admittedly, there are opportunities to lower the cost of providing IT services now, and there may be opportunities for some elements of corporate computing to be delivered as utility services in the future, specifically in providing raw computing power. However, this is by no means the end of corporate computing. This paper considers current trends and discusses the data, software, and control issues that demand the continued long term involvement and responsibility of corporate enterprises within their own IT environments.

Data Issues

Data is not the same as electrical current

To state the obvious, transmitting information is not like transmitting electrical current. I am quite sure that Carr is aware that IT isn't exactly like electricity, but the arguments in the paper clearly encourage this simplistic line of thinking. With electricity, one power 'bit' of 120V AC being transmitted is exactly the same as another power bit and each of these bits could be used by anyone. With IT, one bit is part of a collection of bits that together create a unique piece of information that can't simply be exchanged with someone else's bits. This is a fundamental difference that cannot be oversimplified and this difference raises security issues for corporations that are nonexistent with electricity generation.

Data is a corporate asset providing competitive advantage

Carr's belief that all corporate IT assets will disappear is absurd. The article overlooks the importance of information to a corporation. What is a credit card company without its customer database information? There are few assets more fundamental to a large corporation than the information that it has accumulated – whether that includes the online orders that were processed today, the information on inventory needed in a warehouse tomorrow, or the customer database that has taken 30 years to build.

This wealth of information cannot be ignored as an IT asset, one that also carries an associated responsibility. According to Chris Pratt, Manager - eServer Strategic Initiatives at IBM, for corporations “to go IT free internally they will have to relinquish the tight control they have over their core assets that in many cases they are legally responsible for. (eg a banks and customer data)” (par. 5). The databases and the legal responsibilities associated with the data continue regardless of where the databases may be physically stored. A company still has an obligation to protect the privacy of its customers' information, whether it is stored at a corporate office or in an outsourced data centre. With that responsibility, a person or a group of people at the corporation will be responsible in the same way that a finance department is responsible for the financial aspects of a company despite all of the monetary assets being trusted to financial institutions.

Data Loss at a Utility

Given that data is crucial to corporations, what would happen to the average company if it suddenly lost its data due to an error at a centralized “power” centre? I have known of a few companies that, usually through lack of sufficient backup measures, lost days and weeks of information, some of which was never recovered. Admittedly, these missteps were related to on site issues, but the potential for a centralized provider could prove much more serious.

One utility misstep could potentially harm many thousands of companies at once, creating catastrophic data loss. One assumes that safeguards would be in place to ensure that such an event would be highly unlikely, but highly unlikely and never are two different things. One could argue that the same is true for electrical transmission – all companies supplied by the utility are out of business as long as the crisis lasts. However, with power outages, most companies are back in business when the power is back; with data, the information can be lost forever. According to one source, “[a]mong companies who lose data in a disaster, 50% never re-open and 90% are out of business within two years” (Anderson par. 3).

Data Transmission Speed

Carr also fails to discuss how the needed increase in data transfer speed and reliability for the utility model will be accomplished, except to off-handedly point out that the infrastructure will be provided by network operators. But, will the speed of transmission ever be fast enough?

Akshay Nanduri, Director of Software Development, at Slipstream, a Waterloo, Ontario based company dedicated to providing faster communications for Internet users, believes the transmission issue may never be resolved for two reasons. The first is that “the more bandwidth provided at the network level, the more bandwidth that will be consumed by new applications” and “the transmission medium is a shared resource” (Nanduri par. 3,4).

It is difficult for full featured software applications to run entirely from a remote location because the data transmission speed is not fast enough or is bogged down by other users. For instance, assume that two people are running a standard spreadsheet application from a utility provider, but then 100 people start running the application and then 1000. At some point, the response time on the users' end will become unacceptably slow. Even given fibre optic media and state of the art switching equipment, the laws of physics would still apply.

Incidentally, in a world where speed is often a driver for business, employees who can work faster are more productive, and therefore can provide a competitive advantage. Until utility computing provides the speed of on site computing, few large corporations are going to sign on.
The advancements Carr does mention -- virtualization, grid computing, and web services -- do not address the speed issue. Virtualization and grid computing together provide a means of sharing and generating computing power while web services can provide common tools, and thereby a common user interface for certain applications. Also at issue, however, is that web applications are designed for people to use over the Internet with the speed issue in mind, not in a corporate environment where a speedier, and more full featured interface can be written.

Software Issues

Some, not all, software is utility

Eric Newcomer, chief technology officer at software maker Iona Technologie states “Computers do not work without software. And unlike electricity or other raw technology, software is designed for direct human interaction” (LaMonica par. 10). Admittedly, software exists that might now be described as utility software such as word processors, web browsers, and email. Operating systems such as Unix, Linux, Windows, and Mac O/S allow for these and other software packages to run. However, the importance of software should not be underestimated, and utility software just scratches the surface of what is required in the corporate world.

Software as a competitive advantage

In a world of product overabundance, differentiation is a key strategy for many corporations. If all institutions ran exactly the same way, the software could be identical, but this is simply not the case. Not only do different industries need their own software, different companies typically need their own software. Today and in the future, banks and other financial institutions need to use different software, customized to some degree for each such institution. Software for hotels will not resemble that written for financial institutions and each hotel or hotel chain will likely also customize what it might initially purchase from a software provider.

When purchasing “packaged” software, companies will make decisions regarding how much they are willing and able to make changes to their existing processes in order to use the packaged software. Having worked for several software companies, producing packaged software, there were still requirements from every company for special modifications. These software requirements aren't simply built out of a desire to be different, but out of a need to continue to have a competitive advantage that existed before the software was installed.

The role of software is not to dictate how processes should occur, but to emulate how they do occur. If a manufacturing company uses a process or a set of processes that are distinct from its competitors, allowing it to have a competitive advantage, the software associated with these processes will be customized to match them, not to restrict them. A company may have purchased an ERP software package as its base, but would make modifications, or pay the software provider to make modifications to allow for differentiated processes.

Control Issues

Prioritizing IT

The flexibility of organizations to make decisions for IT based on what is specifically best for them would be compromised in the utility model. With control over IT in the hands of a utility, management at the utility will base decisions on what's best for the utility and its customer base in general, not what may be best for any particular corporation. Some corporations may have little interest or little need to consider such control matters, but others will. For these others, control may be critical.

For instance. there are a myriad of tasks that are done by an internal IT department such as taking the system off line to address a problem, applying patches to an operating system or to utility software. These tasks will continue in a utility environment, but the difference is that the tasks will occur when the utility is ready, not necessarily when its corporate subscribers are ready. Therefore, if a particular patch of an operating system is to take place on systems that are shared, the patch will be 1) scheduled by the utility, and 2) affect all corporations using that set of systems.

In the similar way, what happens when an major upgrade of utility software (ie. a word processor) or of an operating system occurs? Such a change would affect the environment for users at each of those corporations using the utility. For example, the utility schedules an upgrade for all Windows users to move from Windows XP to Windows Vista. The utilities will attempt to ensure that all corporations follow upgrades at the same pace. Why? There is the potential for extra complexity and extra costs required in supporting different hardware and software if different corporations are allowed to use different systems, costs that the utility will attempt to avoid. Alternatively, corporations that resist the upgrades would likely be financially penalized for these extra costs incurred by the utility. In order to continue to save money as expected from using a utility, each corporation will try to manage the upgrade shift on a time frame decided on by the utility, not by the corporation. The set time frame may not match the corporation's resources to allow such a shift, but the priority of the shift has already been dictated by the utility.

Today's Models

Data Centre Model is not for everyone

Carr comments that there are now utility-like services currently being provided by players like Vericenter Inc. in Houston, Texas. This is nothing new. There may be some differences, primarily due to the expansion of the Internet infrastructure, but many outsourced data services have been in business for decades. This author briefly worked for a data centre, HKA Data Processing Corporation in Markham, Ontario that has provided numerous data services from information hosting and backup management to conversions and data recovery for over 20 years (par. 3). Yet, after decades of such "utility providers", we have yet to see the end of corporate computing. Although this structure suits some corporations, providing outsourcing of backup and other IT services, this model does not address all of the IT needs of all corporations.

Private, Flexible Networks Still Required

Carr suggests that in the future thin clients will be used, making “underutilized” PCs unnecessary for corporations in the utility environment. When asked what might hold the thin client, central server structure back from working as a utility, Chris Heaven from beONix Technology, a provider of thin client setups for customers including the Toronto Children Services and the University of Toronto, responded as follows:

The idea is currently viable across private networks where sufficient and reliable bandwidth can be provided. Human resources on-site also provide a measure of protection in mitigating the extent of downtime events when connectivity issues arise.

However, once we move outside of a private network and require external services provided by an ISP a number of considerations on the client side can cause problems; sporadic connectivity, poor latency and insufficient bandwidth can all be show stoppers individually, or collectively.
In my opinion thin clients will not be viable for mass deployments across the Internet until the stability and performance of services provided by ISPs improves dramatically. (par 7-9)

Two issues here are of particular note: human resources, meaning IT staff on site, and a private network, hardly the end of corporate computing. Both measures help to ensure that the thin client setup will work smoothly in an organization. He further notes that thin client devices are discouraged for power users: “[w]ith regards to knowledge workers, who utilize a wide array of applications - many of which are multimedia intense - I would propose that fat client solutions are most applicable” (Heaven par. 20). These knowledge workers access, move and manipulate large blocks of data, resulting in slowing down all of the other workers on the network. It simply makes sense to have a personal system, aka a PC, for these workers to use instead.

Proliferation of Outsourcing, not Utility Computing

Carr's argument for utility computing capitalizes on corporate management's frustrations with IT departments and costs. In the late 1990s, some frustrations stemmed from increased administration costs as cheaper hardware provided a seemingly endless proliferation of servers for individual functions. This wave carried the expectation that different types of services such as email services, DNS services, primary domain controllers, secondary domain controllers, backup domain controllers, backup servers, file services, etc. each required an individual server. Computer rooms were no longer housing one or two larger servers, but numerous small servers, requiring an increase in IT administrative staff.

Then, year 2000 issues cost a lot of money in hardware and software upgrades, all on a short time line, with many believing the upgrades were not necessary and were a complete waste of resources. The animosity between IT staff and management still lingers over who was responsible for not being prepared and the costs involved in ensuring preparedness.

It is not surprising that corporate management began looking at ways to stem this apparent costly, out of control IT environment. Corporations have replaced in house personnel, such as programmers, support staff and call centre personnel, with similar services provided by outside companies, some “offshore”. This transition has been facilitated with the scope of the Internet and the infrastructure built around it, but, by no means, is it an end to corporate computing.

Large corporations still need technical administration staff, even if that staff is being provided by an outsourcing company: someone on site needs to be available for hardware and other administration issues. Large corporations will also need people such as process and business analysts in order to link changes in processes that occur on an ongoing basis with changes in software. There is also a need for coordinating with software providers and sometimes with hardware providers when process changes place new constraints on hardware and other infrastructure.

Final Thoughts

The suggestions presented by Carr outlining the simplicity of the utility computing model are certainly seductive to corporate managers, but that image should be tempered with a healthy dose of reality. While the concept of utility computing may prove to be possible at some level, it is largely wishful thinking. Corporations must continue to be responsible with their databases, and will continue to need flexibility over the software that they use. Although the costs may be lower in a utility computing model, one size simply does not fit all.

IBM once had "Think" as a slogan and this is exactly what management will continue to do to decide what IT strategy matches their needs. And yes, IT is still strategic. Why? Because whenever IT is working well for an organization, it matches what a company does in its processes and its methodologies, as well as its responsibilities. Responsibility for IT management will not go away even if some elements of IT are outsourced or supported and billed like utilities.

The speculation Carr provides is thought-provoking, but I nevertheless remain unconvinced that Carr's concept of utility computing will be realized with such simplicity and similarity to “electrification” in the 20th century. One must look pragmatically at the technological barriers, responsibility issues, and, yes, the complexity of the existing infrastructure before championing a model as the corporate model for the future.

Works Cited

Anderson, Harald. “Data Loss... Can Your Company Survive. (Most Do Not)”.
4hb.com (2004). 19 Sept. 2005 .

Carr, Nicolas. “The End of Corporate Computing”. MIT Sloan Management Review
(Spring 2005): 67-73.

Heaven, Chris. “Re: Thin Clients and the Future of the PC”. E-mail to the author.
22 Sept. 2005.

HKA Data Processing Inc. “Company Profile”. 22 Sept. 2005
.

LaMonica, Martin. Interview with Eric Newcomer. “Will computing flow like electricity”.
CNET News.com. 20 June, 2005. 19 Sept. 2005
.

Nanduri, Akshay. “FW: Short Comment sought”. E-mail to the author. 21 Sept. 2005.

Pratt, Chris. “Re: Short comment request”. E-mail to the author. 27 Sept. 2005.