Technology Corridor

Deck: 
The next big trend is to make network resources more interchangeable and less expensive.
Fortnightly Magazine - March 15 2003


The next big trend is to make network resources more interchangeable and less expensive.

Information technology (IT) infrastructure should be more like the infrastructure that generates electricity, IT consultants say, and at least one Texas utility is listening.

Forrester Research labels its vision for IT infrastructure "Organic IT" and promises that this new kind of infrastructure will save companies big bucks in the long run, without requiring a rip and replace of existing hardware and software.

But how much of this lovely vision is just IT consultant-speak?

Travis Crenshaw, vice-president of IT at TXU, says he has been looking at what Forrester calls Organic IT, and META Group calls adaptive infrastructure, since early 2002. For Crenshaw, this vision of IT infrastructure gives him and TXU the ability to spend the same money on IT, but "to ride it further," even with an increased workload for the department.

"After Texas deregulation, we still have other costs and projects to absorb-for example, a 200 percent growth in server infrastructure alone," he says. Despite such a large spike in work, Crenshaw has added only 20 percent more staff. So instead of outright savings, an organic approach may achieve a cost avoidance. Nonetheless, Crenshaw expects that by 2005, TXU will save 30 percent on IT.

In today's economic climate, it's easy to say that your company doesn't have the money to make dramatic structural or organizational changes, Crenshaw says. But that philosophy doesn't fly with him. People in an organization must be challenged to be creative about using resources, he says. Organic IT is one way his IT department has met the challenge of today's business climate.

In today's economic climate, it's easy to say that your company doesn't have the money to make dramatic structural or organizational changes, Crenshaw says. But that philosophy doesn't fly with him. People in an organization must be challenged to be creative about using resources, he says. Organic IT is one way his IT department has met the challenge of today's business climate.

Restructuring IT

Today's IT infrastructure requires every software application or computer network to have its own individual set of resources, whether servers, storage, or network connections. For example, most utilities use dedicated servers, disk storage, network access, and IT administrators for their PeopleSoft systems, and an entirely different array of those elements for their customer relationship management (CRM) package.

But it's a very expensive way to build an infrastructure. Forrester says large companies typically use approximately 20 percent of their total IT capacity. Imagine how many power plants the country would need if they used only 20 percent capacity for most of their operating time. That, in essence, is the structure of IT in today's large companies. It's akin to having electricity supplied exclusively by distributed generation.

Instead, Forrester says, IT should be structured more like the electricity grid, which spreads the power generation function amongst many plants and re-allocates power generation on the fly, as demand shifts or plants go offline for repairs.

In a nutshell, Forrester argues that within the next five years, IT will see radical overhaul and automation in four key areas: networks, storage, software, and processors. Those changes, the company predicts, will mean vast improvement-and overall lower ownership expense-in three areas:

Utilization. Organic IT will scale up, and down, to match computing demand load, without sudden failures of business capacity. For example, server space could be allocated to accounting during a quarterly closeout, then reallocated to a trading desk during a bid week. Right now, companies must buy enough server capacity to handle peak demand for each software application, and they cannot shift server space between applications.

Integration. Organic IT will quickly and easily connect dissimilar technologies, both within firms and between firms, with the ease of sending e-mail or surfing a Web site. In other words, supply chain management could become easier, because firms could talk to one another without installing a custom integration (see Fig. 1).

Manageability. Organic IT will automate installation, load balancing, failover, and recovery, leaving IT staff to manage unusual events. For example, instead of hovering over server load data every hour of the day, IT staff will be alerted only when there is an unusual breakdown (see Fig. 2).

Several technologies on the market fit into the Organic IT vision, says Frank Gillett, a principal analyst with Forrester. Companies can see significant savings almost immediately from implementing an Organic IT approach, he claims. The key is to focus on two areas that are already organic-compatible: storage and servers.

P.O. Boxes for Computer Files

Currently, storage allocated to an enterprise resource planning (ERP) system can be used only for that ERP system. It's wildly inefficient, Gillett says. Companies buy the storage they think they will need for a specific application for the next three years and then essentially stick those extra storage disks on a shelf until they're needed, he says. But Gillett says storage virtualization technology can change all that by allowing storage to be centralized and managed much more efficiently.

The payoff for moving to storage virtualization could be as high as a 30 percent savings on what companies currently spend on storage, according to Forrester.

Storage virtualization, Gillett explains, is the equivalent of a post office box address for storing data. A file name for locating specific data (e.g., e://users/johndoe/2003 3Q projections) becomes abstracted from the physical location of that data, so that data can be moved around onto different storage disks if overall storage needs dictate that change.

Forrester predicts that leading storage vendors like Network Appliance and EMC will soon deliver networked storage products that will make direct attached disks obsolete. But for now, existing storage architectures like SAN (storage attached network) and NAS (network attached storage) can be virtualized with products from FalconStor Software and Z-force, allowing companies to cancel expensive storage upgrades. (See box for additional Organic IT vendors).

At TXU, Crenshaw and his team have chosen the latter course, using a Hitachi product in conjunction with TXU's SAN environment to pool its storage. That approach has allowed IT to utilize 80 to 90 percent of its storage space, instead of the average company's 50 percent utilization rate. To Crenshaw, it's a good deal: "It's basic economics. [It's better] to be spending $1 and getting a 90 percent return instead of a 50 percent return," he notes.

No More Fiddling With Servers

In today's infrastructure, companies need servers dedicated to Web functions, network functions, particular applications, e-mail, and more. Most servers have a hefty amount of unused space in them-Forrester says on the order of 80 percent, while Crenshaw pegs the rate at closer to 50 percent. Whatever the number, it's not a high optimization rate. What drives such an inefficient, not to mention expensive, use of server resources is the necessity to maintain capacity for peak demand.

As Gillett says, "it's provisioning for maximum usage, even though [the capacity] is only used for a microsecond a year."

Until recently, little could be done to increase server utilization rates. While a company may have had 50 Web servers with a lot of idle capacity, it would take days of "fiddling and diddling" to rearrange software so accounting could borrow the Web servers, Gillett says. Rapid server provisioning promises to change that by automating changes in server configuration, transforming Web servers into accounting servers in mere minutes.

TXU has moved in this direction, with its recent purchase of four Sun Fire 15K servers and a bevy of X-series servers from IBM. The Sun servers are mainframe-class servers, Crenshaw says, but unlike traditional servers, they can be partitioned into various regions. Multiple business applications can run on those servers, he says, meaning that when accounting is preparing for a quarterly close, it gets more space, yet when the trading desk needs more server capacity for a bid week, that same resource can be reallocated there.

By not buying servers that will be underutilized, companies could save hundreds of thousands, if not millions, of dollars. While servers like IBM's X-series servers go for around $500 each, mainframe-class servers like Sun's Fire 15K start at more than $900,000. Putting off the purchase of even one mainframe-class server makes an immediate difference to the IT bottom line.


Articles found on this page are available to Internet subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.