Ask any group of technology managers if they would rather double the speed of their network or double the speed of their PC, and nearly all of them will opt for the faster network.
Clearly, a change in investment priorities toward bandwidth–the capacity of a network to move data, measured by bits per second (bps)–is in the air. Managers are being forced to learn a new, often baffling lingo: T-1 and T-3 lines, fractional T-1, integrated services digital network (ISDN), frame relay, asynchronous transfer mode (ATM), Internet protocol (IP). In this environment, the key constraint on new applications is no longer slow computers, but slow network connections. A high-end PC with nothing more than a 56 Kbps modem is like a race car stuck on a gravel road.
Just as the incredible reductions in the price of computing power over the last two decades have transformed business practices, similar reductions in the cost of bandwidth are beginning to deliver an equally amazing range of benefits. It’s worth noting that every genuinely interesting new computer application in the past several years has had a major network component to it. Bandwidth, in short, now drives the evolution of computing.
Sooner or later, financial managers will encounter the new dynamics of bandwidth supply and demand. An explosion of networked applications will vie for priority in the IT budget, straining a company’s network resources. At the same time, however, an even greater explosion of bandwidth supply means that companies will be able to lower their networking costs, with many more providers to choose from.
And the dynamics of ROI on IT investments will change, too. Companies that upgrade their networks to support a particular application–say, enterprise resource planning (ERP)–may discover additional opportunities to leverage those networks, such as piggybacking voice traffic over an ATM or IP network. (See “Phoning the Future,” Techwatch, February.)
“Demands for bandwidth are growing fairly rapidly, somewhere between 25 percent and 50 percent a year for most enterprises,” says Jay Pultz, vice president and research director for networking technologies at Gartner Group Inc., a Stamford, Connecticut-based information technology consulting firm. New applications such as ERP, Internet and intranet computing, multimedia, and client/server eat up bandwidth, Pultz explains. (For more on bandwidth-hungry applications, see “Next-Generation Applications.”)
Many of these new applications require real-time responses. “Companies want to have a response time across a wide-area network (WAN) as if someone were working [in the office next door],” Pultz says. But getting the levels of performance typical for local- area networks (LANs) on WANs is difficult: wide-area networks are slower by an order of magnitude or more.
“Bandwidth is less a constraint across LANs,” explains Pultz. “You don’t have to buy it per month.” A company can achieve LAN speeds of well over 100 Mbps by installing copper lines, or up to 40 Gbps with fiber-optic lines–more than enough to handle the most bandwidth-intensive application.
“But it’s very difficult to get, say, 100 Mbps on a WAN,” continues Pultz. “If you could, the cost would be prohibitive.” A T-1 line, for instance, that can handle 1.544 Mbps would cost a few thousand dollars a month to support 100 users across the country on a networked app, says Pultz. To go up to 45 Mbps with a T-3 line would cost around $15,000 per month.
With LANs, you also have to factor in hardware costs; switches and routers are needed to direct traffic across the network. Today, companies spend hundreds of thousands of dollars on their backbone (central) LANs, where most of the servers reside. If WAN traffic is increasing by a minimum of 25 percent a year, says Pultz, LAN traffic is growing by a factor of 20, thanks especially to the advent of server-centric, thin-client computing. (See “Commanding the Desktop,” January.) Formerly, 80 percent of a company’s network traffic stayed within small department or workgroup LANs, and 20 percent traveled over the backbone LAN. Today, that ratio is being reversed, as application servers are consolidated on the backbone.v Meanwhile, Internet service providers (ISPs) like MCI Worldcom say demand is skyrocketing on their networks, too, with traffic growing up to 1,000 percent per year on some routes.
But if demand is skyrocketing, so is supply. “I’ve been in the industry for 25 years, and this is the first time that supply will stay ahead of demand,” says Pultz. Major carriers like AT&T, MCI Worldcom, and Sprint offer a plenitude of very high speed fiber-optic connections. Next-generation carriers such as Qwest, IXC, Level 3, Frontier, and Williams have installed tens of thousands of miles of new fiber-optic cabling in the past year. These cables may contain anywhere from 50 to 100 strands of fiber, each strand capable of carrying between 10 and 40 Gbps.
The companies that make fiber-optic transmission equipment for these carriers, such as Lucent Technologies, Nortel Networks, and Ciena, have developed a number of breakthrough technologies that boost the amount of data each fiber can carry. By using multiple fibers, a single transmission system can now comfortably handle data at terabit speeds–that’s 1,000 gigabits, or one million megabits, per second.
The key enabling technology of such enormous capacity is dense wave division multiplexing (DWDM). This technology uses lasers at both ends of a fiber-optic cable to split light beams into a rainbow of infrared colors, each of which can carry data. Capacity explodes: Imagine a fiber-optic cable with 100 strands, each of which can be split into, say, 16 colors–each of which, in turn, can transmit 10 Gbps. Major and next-generation telecom carriers already have thousands of miles of such bandwidth. One carrier boasts that with a DWDM boost, just two strands from its fiber-optic network have enough capacity to handle the networking needs of the entire United States.
As carriers start selling their excess capacity wholesale, to regional Bell operating companies, local exchange carriers, and ISPs, companies will have more choices for cheaper bandwidth. For now, however, most of this excess capacity is going to the top 50 metropolitan areas, says Pultz, with the result that bandwidth will still be fairly hard to get outside of those areas for the next five years or so.
Most companies are still figuring out how best to maximize returns from their network investments. Some take a fairly conservative approach, and focus on reducing costs by centralizing selected IT operations, relocating workers and production facilities to cheaper areas, and improving operating efficiencies. Others focus on developing new revenue, wrapping new services around their existing products, opening new facilities closer to customers, integrating their ordering systems with those of their customers, and transforming their management practices.
We spoke with several companies that appear to be on track.
Sizing Up R/3
Timet Prepares for an ERP System Just as net sales at Titanium Metals Corp. (Timet) zoomed from $185 million in 1995 to $700 million in 1997, thanks mainly to acquisitions, so did the need for bandwidth.
The Denver-based metals manufacturer wanted to update its systems and consolidate the IT operations of its newly acquired units by installing a brand-new, enterprisewide, client/server network–linking sites in the United States, England, Wales, and France. Running on that network, among other applications, would be a thin-client SAP R/3 system (version 3.1h), including modules for finance and control, materials management, production planning, and sales and distribution.
Starting in March 1997, Mark Wallace, vice president of information technology, and network manager Darren Bordeaux began to design a global networking architecture–one providing ample bandwidth for R/3 data to flow smoothly to 900 end-users, without bottlenecks, alongside other network traffic. They had a clean slate to work with. Not only did Timet lack enterprise systems, it lacked networks; the company didn’t even have E-mail until 1998.
The clean slate posed a challenge, too, since the company didn’t have any historical data that could be used to size new networks. How much bandwidth would SAP need? “There’s no hard and accurate formula for what the requirements are,” says Bordeaux. “It’s so business-dependent.” Following discussions with the vendor and similar R/3 customers, Timet developed profiles of user activity and mathematical models to estimate the average per-user bandwidth consumption for the system: 5 Kbps. (To put this value in perspective, modems typically connect at 28 Kbps.)
Conclusion: SAP wouldn’t be bandwidth intensive. Still, as a mission-critical system, it had to be available around the clock. And end-user response time–the time from when an end-user hits “Enter” to when a result is received–had to be acceptably small. Response times of 4 or 5 seconds would irritate users; times of 10 seconds or more might cause system time-outs, with the result that the application could fall apart, says Bordeaux.
The service-level goals Timet set for the R/3 system were 99.5 percent availability on a 24×7 basis, with an average response time of 3 seconds or less. These goals were determined by a cost/benefit analysis that took into account business needs, network architecture, and financial resources. Anything less wouldn’t have been acceptable; anything more would have been unnecessary, and significantly more expensive.
To meet those goals, Timet implemented a WAN using frame relay. Frame relay is a packet-switching technology fast enough for WANs whose traffic can be characterized as “bursty,” or sporadic, yet cheaper than ATM. Each LAN is connected to the WAN by at least one T-1 line. Timet’s Denver headquarters, where R/3 services are centralized, uses eight T-1 lines, with a total capacity of 12 Mbps.
Timet buys the frame-relay service from MCI Worldcom for a total of $1.2 million a year, connecting about 1,250 users in 18 sites across four countries. Subscription to the service varies by location, according to the number of users and the type of network activity. MCI Worldcom guarantees a minimum amount of bandwidth, depending on the network location, at all times.
Last month, Timet went live with its final SAP implementation, in its sites in England and Wales. The company has hit all its service-level goals; average end-user response time is about 1 second. And the actual average bandwidth per R/3 user: 6 Kbps.
Also last month, the company signed a long-term, 24-month contract with MCI Worldcom “granting us significant discounts,” says Bordeaux. “We have an 18-month early-termination clause.” After all, in two years, faster yet affordable technologies might be available. Or, Timet might want to adopt bandwidth-intensive technologies, such as videoconferencing, which would require migrating to a faster ATM network. And that’s a good reason why managers ought to have a basic understanding of their company’s networks, notes Mark Wallace. Before you demand that a favorite new app be installed by Friday, you’d better make sure you’ve got the bandwidth.
Solectron Draws its Partners Up Close and Personal Solectron Corp., a $5 billion Milpitas, California-based provider of electronic manufacturing services, not only has networked all of its worldwide operations, but has extended access
to its information systems to its customers and even its customers’ customers. “Organizations that are geographically distributed need data networks to pull them together,” says Ken Ouchi, Solectron’s chief information officer.
But these systems have to work in real time. One such system Solectron is working on is an order-fulfillment system. If, say, a would-be customer of Ingram Micro (a major PC distributor) logs on to Ingram’s site and wants to buy a machine, the Ingram site will query Solectron’s systems about availability. “They think they are ordering from Ingram, and so delivering the available-to-promise data has to be done in real time,” Ouchi explains.
That kind of ordering dynamic has a number of implications. First, it means that batch-mode EDI transactions can’t cut it. Customers need to know what’s happening right away. Second, it means that so-called intelligent configurators can check the order as it is entered into the system. Rather than having a human take the order and later find that feature x is unavailable on product y as requested, the configurator can ensure orders are entered correctly.
“EDI will never allow us to do all that, so we’re going to the next level above batch transactions,” says Ouchi. “We are building middleware tools that encapsulate business rules.”
But that won’t render the company’s older mainframe and proprietary network-based systems useless. By adding Web interfaces to those systems, the company can extend access to them to employees and customers as needed. Meanwhile, the shift to Web access has also allowed Solectron to move away from proprietary network technology.
In fact, the company WAN is an Internet-based virtual private network (VPN), which, over time, will be used to carry both voice and data. Currently, Solectron uses 17 different ISPs to support its VPN, which is used to transfer information (but not support applications) between the company’s various sites. Ouchi says the company wants to consolidate its VPN with a single large provider that can serve Solectron sites around the world, using its own communications backbone. “ISPs will hand [our traffic] off to the public network,” says Ouchi, resulting in less-efficient communication.
Whichever provider Solectron chooses, Ouchi expects it to share any savings from new technology with Solectron, and to recommend technology changes as appropriate–for instance, when voice over IP becomes much better than voice over frame relay. “We can’t be [expected to be] experts at data communications,” says Ouchi.
Leveraging IT Investments
Alliant Extends New Access to Legacy Systems Alliant Foodservice Inc., a privately held food broker in Deerfield, Illinois, is using its network to centralize distributed operations, and to extend the reach of its internal information
systems to its customers and suppliers. Alliant is one of the biggest institutional food distributors in the country, handling more than 180,000 products.
“For a small increase in networking costs, we’ve achieved enormous savings in computing costs,” says Phil Roszak, director of enterprise architecture at Alliant. “We’re not just offsetting the cost of hardware, but of management as well, because all our IT staff is now in one place.”
Alliant has built its network using fractional T-1 lines, which Roszak says are currently more than adequate to support mainframe access, imaging, centralized E-mail, and other applications. Adding more bandwidth to the network has allowed the company to leverage previous investments in IT. For example, Alliant has been using imaging to make back-office finance applications more efficient for several years. But now, the company has extended access to that system, which was once restricted to a single facility, to the entire corporation.
Alliant has also deployed Web interfaces to its mainframe applications and client/server applications. “We are moving from thick, custom clients to browsers, and that saves us a lot in software maintenance and distribution,” Roszak explains. “We’re starting with financial and human resources applications, and then moving to prolong the life of other legacy systems.”
The company is now working to provide customers and suppliers with access to these systems. “We are looking for opportunities to create a seamless supply chain between our customers and our suppliers, so there will be a lot of other people using our network,” says Roszak. He says Alliant expects to devote a larger percentage of revenues to network services. “Some of the best returns on investment from our capital expenditures have been on IT projects, so we are ramping up spending in this area,” he says. “Our intranet lets us, a distributed organization, standardize practices and processes, and do them more efficiently.”
ROI For Networks
How to Evaluate New Options Warren Buffet recently said that if he were teaching a business class, he would ask students to value an Internet company–then fail anyone foolish enough to give a definite answer. Given that many bandwidth benefits are still the stuff that dreams are made on, how can CFOs evaluate the relative worth of different options?
John A. Meyer, president of the Diversified Financial Services unit of Electronic Data Systems Corp., in Plano, Texas, has some ideas. Meyer notes that advanced networks are bringing about huge operational changes in financial services. Call centers and loan processing used to be totally distributed; now, with electronic networks, they can all be consolidated. Some credit-scoring applications can even be handled remotely with no human involvement, while others can be run from kiosks in shopping centers, connected via videoconferencing to central locations.
Meyer says business managers should be required to determine what such applications might be worth, and technologists asked to calculate how much it would cost to deliver those applications. “That approach lets you rank options with a cost/benefit analysis, and gives everyone some incentive to take a little risk,” Meyer says. “But not too many companies bring that kind of discipline to it. Most [IT organizations] haven’t been in the capital-appropriation business, so they don’t have the same rigor or discipline as those who are.”
John Jendricks, CIO and senior vice president of global marketing at ERP vendor Baan Co., based in Reston, Virginia, suggests investments in bandwidth need to be allocated in a fundamentally new way.
“Traditional IT investments were looked at from the inside out–‘OK, we invest in this system and we can shorten order-to-cash time or increase inventory turns.’ But focusing on internal operations is not the ROI model to look at anymore,” asserts Jendricks. “If that is all you are doing, then you won’t be competitive.”
Instead, managers need to look at investments “from the outside in”–that is, from the perspective of customers and suppliers. “With the Internet, you have to focus on how customers do business with you or would want to, and the real measures are time [saved] and leverage of existing resources,” says Jendricks.
Metcalfe’s Law, named for Ethernet inventor Bob Metcalfe, states that the value of a network grows in proportion to the square of the number of connected devices. Thus, a network with 10 nodes is worth 25 times as much as a 2-node network. Metcalfe’s Law is not exactly a hard quantification, but it’s clear that each additional participant in a network brings value to everyone else already connected.
Jeffrey Ubois writes on business and technology for CFO. Edward Teach is a senior editor at CFO.
———————————————————————— Next-Generation Applications
Increases in CPU speeds have traditionally driven software innovation. Early personal computers couldn’t handle much beyond text: graphical interfaces came later, as speeds improved.
In the same way, increases in bandwidth have driven innovation in network applications. Not long ago, network apps were almost entirely text based. Image, audio, and video files simply took too long to download routinely. With sufficient increases in available bandwidth, the Web was born, and today there is promise of a host of new applications.
Predicting the next killer network app is usually an exercise in hubris. But it’s possible to chart the limits of the possible. Among the areas worth considering:
Remote collaboration. Videoconferencing, audioconferencing, whiteboarding, and other real-time interactions are just slightly beyond the capabilities of today’s Internet. But not for long. So Solectron Corp.’s CIO, Ken Ouchi, is particularly interested in applications that support remote collaboration. “No matter how fast I fly, if I want to be in Europe or Asia for a meeting, I burn a day and a half to get there,” says Ouchi. “It’s not just dollars, but time.”
Voice over IP. This is another area where companies can begin to reap quick returns from network investments. Today, voice accounts for close to three-quarters of Solectron’s telecom budget, but Ouchi expects that to change. “We spend much more on voice than on data, but when my voice, video, and data are all on the same [IP] network, in the future my telecom bill will look more like my Internet bill.” That means flat-rate and cheap.
Remote imaging. Imaging isn’t new, but the ability to easily access image databases from anywhere is only now becoming common. Companies like Electronic Data Systems Corp. (EDS) find that extremely valuable. “We have four remittance shops for our customers in the U.S., so if someone is operating nationwide, they can send bills to their California customers via the L.A. site rather through Memphis, which eliminates the float time and gives them real money in their pocket sooner,” says John A. Meyer, president of EDS’s Diversified Financial Services unit. “We can also consolidate operations through imaging by centralizing research and exceptions processing.”
Education and training. Just-in-time delivery of knowledge and training is another natural application for high-speed networks. Many universities are now making classes available on the Internet; companies may also want to consider making their training materials available via the Web.
Customer support. In some sense, the best customer service is face-to-face. But face-to-face isn’t available on demand, and it isn’t cheap. So companies are looking for Web-based substitutes. For example, Baan Co.’s Cyber Consult Service is using ActiveTouch Inc.’s ActiveMeetings service, a Web-based conferencing system, to solve customer problems in cyberspace.
Supply-chain management. ERP works only when fast networks are in place. But the payoffs can be huge. “One of our customers has a business that is basically cash management and materials management,” notes John Jendricks, CIO and senior vice president at Baan. “They have thin margins.” But now, he says, the company’s margins are widening, because it’s “moving to central servers that give everyone real-time access to their cash position, orders outstanding, and their materials position. It wouldn’t have been possible without their WAN.” Jeffrey Ubois