Rationalizing Tech: When Less is More
By Sam Collier and Steven Marks
LARGE firm I.T. departments struggle to manage an ever-growing array of applications and servers distributed in offices worldwide. More servers mean more management, higher upgrade costs, and of course, more things to go wrong.
The world of "distributed computing" envisioned 10 years ago never contemplated either the large number of systems and applications, or the dramatic shortening in product upgrade cycles of today. To many, the costs (financial, operational and quality) of distributed computer systems are beginning to outweigh the benefits.
Until recently, the idea of consolidating major systems under one roof remained an unexplored dream. Now, more firms are seeing the possibilities of system consolidation: taking servers and systems from far flung offices and centrally co-locating them with the senior I.T. staff.
"We were drowning in servers," says Brad Christmas, C.I.O. of Akin Gump, Strauss Hauer and Feld L.L.P. "We have 14 offices around the world and 157 servers. Now we are looking at bringing our server count down to 27. That buys us a lot."
Fewer servers mean fewer parts to purchase, patch, upgrade and maintain. It also means fewer parts to break down. Fewer servers allows engineers to invest money into more redundant systems, and server clusters. Centralizing servers also maximizes expensive systems, like storage area networks (SANs) and tape libraries.
Consolidation also enables an economy of skill, where highly skilled staffs are centralized reducing redundant functions. Economies of scale are good; economies of scale and skill are better. The proof is in the results: lowered hardware costs, reduced staff and training costs, fewer systems outages and faster, cheaper upgrade cycles.
A new technology, "pure optical" IP bandwidth, offers significantly lower bandwidth unit costs. That presents the opportunity to redesign legal information system networks. Like the microprocessor and the hard disk, and even the VCR, the lower the price, the more people can (and will) use.
Perhaps the easiest way to understand it is perhaps by asking a simple question: "If you could purchase as much WAN bandwidth as you wanted for just a few hundred dollars per megabyte per month, anywhere in the world, how much would you buy?"
The answer is a lot, with the opportunity to totally change the architecture of modern legal networks.
Faster and cheaper are always important, but quality of service is still critical. The dirty little secret in WAN technology has always been that while frame relay, ATM, T1s, E1s and T3s sound fancy; all wide area networks have run over older voice telephone technology. By today's standards these older voice networks are slow and expensive. By tomorrow's standards they are ancient history.
Turning a WAN into a LAN
Next generation 'pure optical' carriers have created completely new networks from the ground up. These networks were designed purely as high-speed TCP/IP data networks without the problems inherent in legacy telephone systems.
The basis of these networks are pure optical "wave division multiplexed" technology, using new protocols such as Multiprotocol Label Switching (MPLS). Switching from running wide area networks over voice telephone networks to a true data network makes a vast difference in how networks work. High speed networks hold the potential to virtualize the WAN, turning it into an extension of the in-building LAN without the performance penalty or scalability limit inherent in legacy services (such as frame relay) built on SONET voice architectures.
Companies such as Global Crossing, Level 3, Broadwing and XO have spent the past few years building these new networks worldwide. While the AT&T's of the world continue to enjoy the predictable cash flows from voice toll traffic, the growth in demand for data traffic will far outstrip voice especially as voice over IP turns more and more expensive long distance minutes into cheap TCP/IP data packets.
In the past, WAN costs were based on a combination of speed and distance. Next-generation networks have very high capacity scalable backbones with dramatically lower operating costs spanning virtually all major metropolitan markets, making location largely irrelevant. The resulting collapse in per-MB bandwidth costs, coupled with operationally efficient pure optical networks is creating a distance-insensitive pricing model for connectivity. For example it is possible to get a long haul IP connection (of up to a gigabit in capacity) for about $300 per MB per month in all of the major U.S. metropolitan areas.
Consolidation adds the fundamental risk of placing all of your eggs in the same basket. Yet, this risk is more easily mitigated than the existing risks in current network designs. Most network implementations currently have a high level of risk ranging from inadequate hardware, under qualified data centers, lack of backups and an inability to integrate redundant systems because of the absence of economic scale in many locations.
Really reducing risk to an acceptable level requires architecting multi-vendor networks to provide redundancy in case of a failure, which has proved too expensive using current telephony based WANS.
Mitigating local loop issues, using new technologies being delivered by metropolitan area network providers such as Yipes, Sphera, Intellispace and Cogent, is another major component to creating a reliable next-generation data network.
"At first we were concerned about increasing our risk," says Christmas. "But as the design developed, we were able to build in a level of redundancy that greatly increases our ability to survive service outages. There is a false sense of security having servers in every office."
A New Equation
Switching to a modern data network will not save you money directly. Done right, such a network could cost significantly more. However, the higher speed, higher quality WAN enables other projects, such as server consolidation and development of true business continuity systems, where the cost savings more than make up for the additional WAN cost as well as creating a more manageable and reliable network.
In the end, the ability to centralize and 'right size' services is the key to delivering real business flexibility and provides for a better business continuity model than exists today.
This new technology architecture enables firms to increase technology service levels, especially to branch offices, without the escalating increases in cost.
Like most major technology trends, this one is not happening overnight. But just as the Internet existed for many years before becoming ubiquitous, the notion of substituting bandwidth in place of distributing redundant systems in every office, is starting to mature and will likely accelerate.
Sam Collier is a principal of Union Square Technology Group and president of Aspire.net. Steven Marks is executive vice president of Aspire.net.