Before delving further into the details of the Next-generation Enterprise WAN architecture I think it will be instructive to look first at the history of the Enterprise WAN.
Some would argue that the X.25 networks – the first "cloud" services in widespread use – that were popular in the 1970s and 1980s for remote mainframe terminal access are the beginnings of the Enterprise WAN, but I’ll stick to the "modern" era that began with the rise of the personal computers connected via LANs in the 1980s.
Back then, if you wanted to connect LANs that were not in the same location, you used point-to-point leased lines. These were typically DS0 (56 Kbps!) connections, and then the more expensive T1/E1 or T3/E3 connections, and given their significant expense, fractional T1 or T3 lines as well. The connections were first done using remote bridges at each end, and later with devices called routers, popularized by this little company you may have heard of named cisco…
At the beginning of the 1990s, Frame Relay service was introduced. While using the same DS0 and fractional or full T1/E1 and T3/E3 connections, by connecting to a "cloud" from a service provider, you no longer needed to purchase and manage individual links between each of the locations that you wanted to connect. Frame Relay service offered much lower monthly WAN costs, far fewer physical connections to manage, allowed the expensive last-mile link bandwidth to be shared (and thus used more efficiently) across multiple remote connections, and required less expensive router hardware than the point-to-point alternative. This overwhelmingly compelling economic story – OpEx and CapEx both – caused a revolution in the corporate WAN and contributed to its rapid growth. Within 5 years of its introduction, even the most conservative enterprises such as banks had migrated to Frame Relay. It was the fastest uptake of any WAN service in history – including the public Internet.
Those who remember their history know that ATM services were introduced in the 1990s as well, but they never achieved any meaningful enterprise WAN market share.
Multiprotocol Label Switching (MPLS) is the successor to Frame Relay. The word "multiprotocol" in the description notwithstanding, it was designed as an IP-based solution for carriers to converge voice, video and data on the same network. The uptake of MPLS began in the middle of the 2000s, and by today, while Frame Relay service continues to be available, the clear majority of enterprises have migrated their WANs to MPLS in the last few years. The carriers have generally priced MPLS slightly lower than Frame Relay on a price/bit basis, and have made it clear to their customers that Frame Relay is being phased out in favor of MPLS.
While Frame Relay is a connection-oriented, circuit-switched protocol, and MPLS is a connectionless protocol, from the enterprise’s point of view, the technical differences are not substantial. The major exception was for VoIP, where the connectionless nature of MPLS enables any-site connectivity useful for reducing latency in live voice calls, and where MPLS has a practical edge in delivering QoS.
The Internet, of course, has a much longer history, but it began to "happen" for enterprises after the 1994 release of the Netscape Navigator web browser. Enterprises bought T1 and later T3 connections from Internet Service Providers (ISPs) for connectivity.
IPSec VPNs, introduced in the mid-1990s and popularized in the late ’90s, made site-site connections over the Internet secure. But despite the price advantages of Internet connections, and despite the fact that the Internet in the last 15 years has revolutionized just about everything else that touches IT, very few large enterprises today use the public Internet for their site-site intranet connectivity. Of course, today almost all enterprises do use it for individual home and mobile remote access –which was accelerated with the introduction of SSL VPNs – and many use it for backup connectivity when the primary MPLS link fails.
As noted in my last column, MPLS today remains very expensive compared to public Internet connections, anywhere from 20 to 100 times the price/bit. In future columns, we’ll discuss in more detail why this is so, and why it’s likely to remain the case. Frame Relay, which was the price/performance leader when introduced in the early 1990s, barely went down in price from about 1998 to 2003, and both Frame Relay and MPLS have come down about 15% to 20% per Mbps per month each year for the last several years, but this doesn’t come close to the roughly 40% annual reductions in the price of Internet bandwidth. When IT budgets were expanding rapidly over a decade ago, additional yearly spending was manageable if it was used to help WAN service keep up with data traffic, which has continually followed Moore’s Law and grown 40% or more each year. With much flatter IT budgets in the last several years, however, this has been a problem, and thin WAN pipes are frequent bottlenecks for network application performance.
The high cost of and relatively low bandwidth available with Frame Relay and MPLS at most locations are just some of the reasons WAN Optimization technology became popular in the 2000s (though certainly not the only, or even any longer the main, reasons WAN Opt is so widely deployed today).
Broadband Internet connections – ADSL and cable modem – came along beginning in the late 1990s to enable nearly universal high-speed Internet access, at much lower cost than T1-based access. As a result of the continuing increases in bandwidth available with these technologies, and more recently with 4G/LTE as well, most home users and many mobile users have higher-speed Internet access at their homes than do workers at smaller sites in the vast majority of enterprises.
Enterprises, especially large enterprises, typically did not deploy these broadband connections at their own locations at first, although over the last several years many now use them at branch offices for local access to the Internet or for the aforementioned backup VPN connections.
Despite this very large, and growing, price/bit advantage, enterprises stayed with first Frame Relay and now MPLS because the public Internet on its own is not reliable enough. By reliability, I don’t mean simply connection availability, but specifically the predictability of performance.
On the Internet, there is no single service provider guaranteeing end-end performance. You can’t get an end-end SLA over the public Internet. The business model of peering points – the places where the network of networks that is the Internet connects different ISPs to each other – specifically precludes the ability to deliver any such SLA. You simply can’t get meaningful QoS over a single Internet connection, both because of the generic peering point use and the difficulty of doing the accounting or the billing even if there were service providers who wanted to try.
And the above even applies when the last mile is a TDM connection such as T1 or T3. With inexpensive broadband links, there are the additional problems of lack of sufficient upstream bandwidth on ADSL, the lower MTBF and also the much higher MTTR associated with broadband connections which have been optimized for the mass market of homes and SMBs, rather than large enterprises.
And thus is the state of the Enterprise WAN until very recently: expensive but reliable Frame Relay and now MPLS connections with relatively little bandwidth at remote sites, often with IPSec VPN backup connections and frequently with WAN Optimization to help squeeze more capacity and performance from the slow expensive links, coupled with cheap and getting cheaper but still "unreliable" Internet connections at data centers and larger sites for both generic Internet access and remote user connectivity.
With the challenge of private and public cloud computing facing enterprise WAN managers, combined with the continuing inexorable growth in file sizes and the demands of new applications, the situation is ripe for a NEW architecture to revolutionize the enterprise WAN at least as much as Frame Relay did more than 15 years ago.
A leading expert in WAN/LAN switching and routing, 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.