Storewidth is the combination of ubiquitous, high-capacity optical bandwidth and unlimited storage capacity, but we're not there yet.
BY JON OLTSIK
In his book, Telecosm: How Infinite Bandwidth Will Revolutionize Our World, George Gilder dedicates a chapter to "The Storewidth Paradigm," which he defines as follows: "Storewidth is the conversion of abundant bandwidth and heterogeneous petabytes into accessible information. Accessibility, in turn, depends upon the inexorable light-speed delays that afflict the Internet. ...In communications, the speed limit is the light-speed wall-latency-and the only way to mitigate it is to move the data closer to you. When latency rules and enforces a tyranny of the initial bit, bandwidth gives way, as the governing principle of the network, to storewidth. Delay stems from the sum of search time and delivery time."
In other words, storewidth is information nirvana. It's the aggregation of ubiquitous, high-capacity network bandwidth with equally available and plentiful storage capacity. When we get there, it will be spectacular, but first there are some obstacles to overcome. Regardless of increases in network throughput, there will always be a need to intelligently move information around the network to bridge the latency gap. Why? Simply put, latency equals time, and time equals money.
The storewidth paradigm assumes that the Internet-which provides ubiquitous, abundant optical bandwidth based on inexpensive Ethernet technology-will become a global highway for exchanging a wealth of digital objects and services in near real-time. While storewidth is a compelling vision, several obstacles prevent it from becoming a reality today:
Metropolitan fiber is still scarce. A fundamental premise of storewidth is the availability of infinite, optical-network-based bandwidth. How ever, while companies like Global Crossing, Level 3, and Qwest blanket the world with high-capacity, long-haul optical pipes, metropolitan area fiber deployment is still in its infancy. Over time, massive network construction efforts will address the dearth of fiber in the ground, but for the immediate future, low-tech issues like right-of-way negotiations, building access, union contracts, recalcitrant landlords, and even bad weather will continue to slow metro area fiber build-outs.
Storage is still expensive. As with network bandwidth, the storewidth vision is incomplete without cheap, abundant storage capacity. While the average cost per megabyte of storage continues to decrease by approximately 30% per year, enterprise storage requirements are growing at two to three times that rate, meaning that overall storage spending continues to rise. In the Global 250 market, moreover, prices have yet to come down dramatically: High-end subsystems from vendors such as EMC, Hitachi, and IBM easily fetch more than $1 million.
Large companies use private networks-not the Internet. The storewidth vision dictates that the Internet will be the network of choice for universal data movement, but today's large businesses run on a rat's nest of expensive private circuits, legacy networks, and rigid data services like Electronic Data Interchange (EDI). Why? The Internet is still fundamentally a shared, slow network that is too unstable and non-secure for most corporate traffic. It will take years of improvements and experience before most CIOs will move "bet-the-business" traffic over the public Internet.
Complex content is in its infancy. The storewidth concept also hypothesizes that networks ultimately become superhighways for the movement of huge rich media files. Despite industry hype, however, large enterprise customers currently devote little time or money to multimedia content like streaming audio and video. Multimedia applications languish in part because economic conditions preclude new projects. New e-business application development efforts labeled "gotta have" in 1999 have been indefinitely postponed in the harsher economic climate of 2001. Enterprise IT efforts now focus on streamlining processes, maintaining legacy applications, and cutting costs-not building multimedia applications with an often-fuzzy return on investment.
Also, production problems persist. Multi media application development is as much about professional production skills as it is about technology. Leading companies like General Electric, Ford, and Citigroup will not release content on the Web unless it meets their strict quality guidelines-an expensive and time- consuming task.
Ethernet shares the stage with many other protocols. Another storewidth assumption is that all network and storage technologies adhere to a common protocol, namely Ethernet (and IP). There are many technical and economic reasons why this vision may one day come true, but today's enterprise metropolitan networks are a patchwork of Ethernet, ATM, frame relay, ESCON, and Fibre Channel. In the near- to mid-term, many data networking protocols will converge to Ethernet, but storage remains-and will be for some time-tied to transaction-friendly channel protocols.
The good news
The gap between today's enterprise networking realities and the storewidth paradigm might make it easy to dismiss it as a mere vision. While many of Gilder's assumptions may be years away, the concept already has a sound foundation because of these four enablers:
Enterprise-class service providers bridge the fiber gap. A new class of enterprise-class service provider is helping enterprise IT organizations overcome the realities of today's dark-fiber infrastructure. For example, this type of service provider helps large businesses overcome the metro fiber shortage by leveraging core expertise in finding fiber routes, mapping fiber provider facilities, procuring capacity, and building laterals and risers. By employing these skills on behalf of customers and establishing relationships with dozens of fiber providers, these service providers effectively remove the frustrations of fiber procurement from enterprise customers.
Optical networks act as a high-speed gateway between private and public networks. A new, conceptual network architecture called enterprise optical networking will create a light-speed link between today's private enterprise networks and the Internet-centric storewidth vision. Think of enterprise optical networking as a high-speed, protocol-independent, optical metro backbone. With this infrastructure in place, companies have their own light-speed highway to move network and storage traffic from facility to facility. Transferring traffic to the Internet is a simple matter of connecting these private optical nets to ISPs and Internet peering points. Through this combination, enterprises get the best of both worlds: a high-speed metro infrastructure for internal communications and a bridge to the public Internet as it used to transport more and more enterprise data.
Data movement applications traverse metro-optical networks. While the rich media applications described in the storewidth vision have yet to take hold in the enterprise, it doesn't mean that other large-scale data movement applications are not being deployed. Rather than using training videos, however, these applications include remote mirroring, system backup, and database replication. With storage requirements growing at approximately 100% per year, these data movement applications will be the primary storewidth driver in the metro enterprise networking market.
Dense wavelength-division multiplexing (DWDM) technology provides a fast pipe for all traffic. Enterprises with metropolitan networking requirements don't need to wait for protocol unification around Ethernet. DWDM technology from vendors like Cisco, Nortel, ONI, and Sycamore can accommodate legacy protocols like ATM, ESCON, and Fibre Channel, as well as Ethernet. Support for these protocols enables customers to use optical technology today and offers investment protection against future technical directions like storage over Ethernet (e.g., IP Storage and iSCSI) or new protocols like InfiniBand.
Enterprise optical networking provides the fundamental building blocks to transform the storewidth concept from vision to reality in the enterprise networking market. Large companies already move large amounts of increasingly time-sensitive data over optical networks today, and as enterprise optical networks proliferate, data-movement intelligence will move to the network. Enterprise IT is a sea of homegrown and third-party data movement applications living on a plethora of proprietary boxes. By 2004, this model will become extinct, as the network will control data-movement policy, priority, routing, and security. Applications like replication, backup, content delivery, and software distribution will be written to the network through open APIs, communicate via XML, and use LDAP directories for authentication and access control. By using these open standards, enterprises will be able to plug and unplug applications and rich content with relative ease.
Service provider business models will take off and become intertwined. Once an enterprise optical networking infrastructure is in place, a slew of network-based services will have a direct path to enterprise customers through business relationships with a new type of enterprise-class service providers and connection points at network points of presence and carrier hotels. This new high-speed access will accelerate the adoption of an outsourced Web-hosting market and will finally make the storage service provider (SSP) approach credible. To provide enterprise-class quality and care, service providers will work together to address market needs, integrate management tools, and exchange customer information to provide true end-to-end service level agreements.
A greater reliance on SSPs won't be the only effect of enterprise storewidth on the storage market. By moving storage intelligence to the network and away from proprietary systems and software, users will leverage a slew of IP features, network software, and applications to gain more efficient, and ultimately cheaper, options for data movement, protection, mirroring, and backup.
Going forward, DWDM and CWDM (coarse wavelength-division multiplexing)-based enterprise optical networks will have ample bandwidth to support any multimedia requirement. Most applications won't come from in-house efforts; rather, they will be provided by established content providers like Bloomberg, Reuters, and Dow Jones. Expect every trader on Wall Street to have real-time video on the desktop by 2003.
The emerging enterprise optical networking market validates George Gilder's storewidth concept, as both involve massive data movement over an optical network infrastructure. As enterprise-class service providers deliver enterprise optical networking solutions to business customers, the enterprise storewidth vision will become a reality, providing metro-based companies with business advantages through new types of network intelligence, applications, and service options.
Jon Oltsik is vice president, marketing and strategy, at GiantLoop Network (www.giantloop.com) in Boston, MA.