The three eras of networking

For the past several years, businesses limited their investment in networking infrastructure. This is now giving way to new growth and vibrancy as we enter a third new era in the networking industry’s 35-year history.

The mainframe era
The first era, starting in the 1970s, was the mainframe-centric model of computing and communications. Networks were hierarchical and used dumb terminals. Network connections were slow, costly, complex and by today’s standards, unreliable. Users numbered in the thousands and the dominant standard was SNA – standing for systems network architecture – an IBM proprietary architecture.

The connectivity era
The connectivity era, began in the early 1980s with the widespread adoption of Ethernet (invented by Bob Metcalfe, who would later become 3Com’s founder). Ethernet initially connected PC workgroups sharing files, printers and applications. Over time millions of local area networks came into being to support this broad sharing of resources. As the 1980s progressed, developing wireless area network (WAN) technologies interconnected millions LANs. The internet, or more accurately, Internet Protocol (IP), then enabled a global network of networks. Second era networks were much faster, more reliable and easier to use than earlier networks, and the number of users expanded into the billions.

While these data networks proliferated, other networks were also in development – being purpose built to handle specific traffic types, including voice and video.

Era of convergence
The third era of networking will ultimately be known as the Era of Convergence. Now that all network traffic is in digital packets, it is possible to create single networks to handle all types of traffic.

There are several compelling reasons to build converged networks. A significant benefit is that they deliver richer, more compelling applications and services to end-users. These new applications can integrate voice, data, and video in unique and productive ways offering a richness of experience not possible in many of today’s applications. Another major result is the potential to dramatically lower costs by reducing infrastructure and staff. As converged networked tend to carry more internal traffic, they also tend to reduce costs payable to service providers.

But what are the challenges? To understand them, it is helpful to know how data (or IP) networks operate. Today’s networks open packets to find intended destinations, send packets, confirm their arrivals, and forget they ever saw those packets. Open, send, confirm, and forget. That’s it. It’s ‘best efforts’ delivery that handles packets without distinguishing between different applications and different users.

In today’s data networks, packets lost in transmission must be reconstructed, which is fine. It’s not too much of a problem if network capacity gets tight here or there, or if too many hops delay some packets. While some data traffic is mission critical, much of it isn’t. As a result, such networks don’t distinguish between data and other traffic – they simply open, send, confirm and forget.

In converged networks however, voice and video must arrive in real time. That is, there can be no tolerance for jitter, latency or down time. If a data packet is lost and reconstructed, no one cares. In a data world delays are not noticed. In a voice and video world, it is unacceptable to lose and reconstruct packets.

Era of intelligence
The point is that future networks must do more than open, send, confirm and forget. They must open and fully inspect packet contents. They must compare what they learn about each packet with a specific set of instructions, and potentially with thousands of instructions, to make informed decisions about how best to handle each packet; all at line speed.

Let me present some examples. A network opens a packet and asks, “What application is this user opening?” It turns out to be a voice packet. So the network automatically prioritises that packet to more efficiently traverse the network. The network might even route the packet a certain way to achieve lowest latency or lowest cost. The network might ask, “What application is this?” and send it to the least utilised server. It could ask, “Is this user authorised to be on the network? Is this user allowed to access this application?” It may then authorise access as appropriate. The network might ask, “Is this packet infected? Is the flow malicious or anomalous?” Depending on the information it receives, the network may instantly quarantine packets with the potential to infect your network or applications.

Each of these are examples of Total Packet Inspection – moving past “open, send, confirm and forget” to understand everything about packets and traffic flows, applying that information against a set of instructions, and having the network make decisions automatically. This is the era of Intelligent Networks.

Most examples above deal with security. While an intelligent network can address more than security, there is no more pressing need today than to apply Intelligent Networks to security problems.

The coming security deluge
As the world moves to converged networks of data, voice and video, the number and type of security threats and incidents will multiply beyond the scope of today’s threats.

The main reason for this has to do with how people connect with enterprise networks. In the past, network access devices mostly stayed inside the physical boundaries of the enterprise – most of these devices were PCs. This is no longer the case. Most devices attached to the network today are portable; they work at home, on the road and in the office. They perform personal as well as business functions. Because of this, they are more vulnerable to infections being brought into your network. In fact, network security risks from inside the enterprise may be greater than the risk of attack from outside. This problem is worse in converged environments where the number, type and mobility of devices accessing the network include a variety of wired and wireless devices (for example, Blackberry and Palm units).

Industry standards or proprietary technology?
Whether converged networks use products based on industry standards or proprietary technology is an important question for the chief information officer. The answer is that open architectures and interoperability are not “nice to haves” but “must haves”. Why? Enterprise need not spend millions of dollars replacing infrastructure to meet the security demands of converged networks. All that is needed is a layer of intelligence atop existing infrastructure to handle security and convergence. This will allow network intelligence to be added when and where it is needed, without forklift upgrades. This will enable the market to buy the best technologies at competitive prices.

Conclusion
Convergence is a reality. The business advantages of single networks carrying all types of traffic are compelling. But the promise of convergence requires a move into a new era – the era of intelligent networks. In that era, networks will be capable of deep and total packet inspection to keep infrastructure and applications secure as well as operating efficiently. Networks will automatically prioritise and secure applications in conformity with user and business policy requirements.