Disrupting the high-margin WAN services business model

It usually takes years, if not decades, for newer, “cheaper/faster” technologies to completely replace the older ones. After all, we still have mainframes more than 35 years after the introduction of the PC. It took more than 15 years for PC and LANs and “client-server” to essentially completely replace minicomputers, and about the same amount of time for LANs, WANs, TCP/IP and routing to replace FEPs and SNA.

One notable exception to this was when RAID revolutionized the storage industry in the late 1980s and early 1990s.

As I noted when initiating this column about the Next-generation Enterprise WAN (NEW) architecture, the key technology catalyst, WAN Virtualization, does for the WAN what RAID did for storage. The analogy works both at the business level and at the technical level. We’ll look primarily at the business benefits here; we’ll cover the technical parallels in a future column.

Before RAID, in the 1970s and the 1980s there was a large market for expensive mainframe-attached storage systems, which IBM dominated, and similar markets for minicomputer storage. IBM’s margins were extremely high, and even the vendors of “plug compatible” storage systems had high gross margins as well. They competed on the benefits of their proprietary disk-drive technology, and only within a narrow range – a factor of 2 or 2 ½ times at most – on cost.

Along came Seagate 5.25-inch hard disk technology, sold at first to the nascent but quickly growing market for personal computers. Each hard disk was not nearly as reliable, having nowhere near the MTBF or seek times of the mainframe/mini disks, but were built in high volumes with enormously lower price/bit.

RAID – Redundant Array of Inexpensive Disks – took advantage of that PC hard disk price/bit to revolutionize the enterprise storage market. RAID solved the reliability and capacity limitations of the PC hard disk, and enabled them to be leveraged for data center, business-quality storage systems. With RAID technology, companies like EMC built bigger, faster, cheaper and more reliable storage subsystems. Within a very short time, no one tried to build big, fast, reliable proprietary single disk storage systems any longer.

For WAN Virtualization and the NEW architecture, the existing high-margin market is that for private WANs, which today means MPLS. Where IBM was the single dominant storage vendor before RAID, today with private WANs, the oligopoly of AT&T and Verizon (and to a lesser extent, Sprint) in the United States, and government-sponsored telecom monopolies or near monopolies in other countries dominate the market, with one or two vendors in each company controlling 70%+ market share for these services.

Pricing of private WAN services, which was somewhat aggressive and clearly related to the cost of the service when Frame Relay was introduced in the early 1990s, now bears no correlation whatsoever to the cost of providing the service. Rather, due to the last mile access monopoly by the RBOCs (for copper connections and frequently fiber, too) and this oligopoly of nationwide / global private WAN service providers, the price/bit of these services has only slowly come down in the last decade, even as the price of almost everything else in technology has followed the classic Moore’s Law curve downward.

The irony here is that the enabling technology for WAN Virtualization that is analogous to the PC hard disk for RAID is public Internet services – many of which are provided by these very same service providers! Just as with the PC and the PC hard disk in the 1980s, massive investment in the Internet and broadband has been targeted at individual consumers and small business, and given the combination of high volumes and competition in that market, the price/bit there has continued to come down with Moore’s Law. Internet access now costs $1.50 – $15 per Mbps per month, versus the $300 -$600 per Mbps per month that the telecom SPs charge for MPLS.

While WAN Virtualization technology is fairly sophisticated in terms of how it makes real-time per packet measurements and sub-second response, just as with RAID, the price/bit gains come from providing the surrounding reliability to enable leverage of Internet economics. Where RAID wrapped a layer of hardware and intelligent software around multiple PC hard disks, WAN Virtualization’s two-ended, appliance-based solution does something similar with multiple WAN connections – existing private WANs and high-speed Internet connections (T3, OC3, Metro Ethernet, etc.) at data centers, large sites and colocation facilities, as well as existing connections and any type of broadband Internet links for branches or smaller locations.

RAID leveraged PC hard disk technology to revolutionize business storage cost, capacity and reliability. WAN Virtualization leverages the most powerful, ubiquitous, low-cost communications network ever created – the public Internet – to deliver Enterprise WANs that are far higher bandwidth, far lower cost and more reliable than the best proprietary single service provider WANs available today.

As documented in places like Clayton Christensen’s book The Innovator’s Dilemma, it usually takes longer for new technologies with better economics to be fully embraced by the mainstream because of either limited capacity or especially lower reliability. RAID was the rare exception and quickly replaced single-disk enterprise storage solutions because it actually delivered greater reliability than those systems from the beginning, and because it could leverage multiple disks, quickly overcame in capacity and performance as well, while remaining fully compatible with existing computing and networking systems and so causing no disruption to the computing environment in order to be implemented.

WAN Virtualization does the same thing for enterprise WANs in terms of reliability, delivering a network with higher reliability and greater application performance predictability even when using only lower-quality Internet connections. The use of colocation facilities further improves the economics, network diversity and performance. WAN Virtualization actually delivers even smoother migration than RAID did, because you can include existing private MPLS WANs as part of the mix, and deploy it piecemeal in incremental fashion, capping expensive private WAN spending, and removing MPLS later when and if desired. Capacity limitations – the best WAN Virtualization solutions top out at about 1 Gbps today – mean that it’s not designed for the highest-end data center-to-data center connectivity, but that’s the main limitation today, and that limitation doesn’t preclude its use for connecting smaller locations to data centers, or for use with the vast majority of data centers that have less than 1 Gbps WAN bandwidth today, in the same way that WAN Optimization has been successfully deployed over the last several years.

For good reason, most enterprise IT folks dislike “two steps forwards, one step back” technology deployments because of the disruptive effect they can have on operations, or simply the risk thereof, which is why so many technologies even with excellent price/performance characteristics can take years before being fully adopted. Those that offer those “two steps forward” in terms of economics and OpEx/CapEx savings without requiring forklift upgrades or sacrifices in terms of overall system reliability can be adopted much more quickly.

It may well take a bit longer for WAN Virtualization and the NEW architecture to change the wide area networking world as completely as RAID changed enterprise storage, but the parallels in terms of surrounding superior economics with business-class reliability are striking.

A twenty-five year data networking veteran, Andy founded Talari Networks, a pioneer in WAN Virtualization technology, and served as its first CEO. Andy is the author of an upcoming book on Next-generation Enterprise WANs.