Mon, 26 May 2003

Thirty years on, networks are getting smarter

Zatni Arbi, Columnist, Jakarta,

One of the components in our desktop and notebook computers that we increasingly cannot do without is the Ethernet card. A more appropriate name is perhaps the Network Interface Card (NIC); however, if you were to go into a computer store in Mangga Dua, Jakarta, and ask for a NIC, the store attendants might give you a blank look. On the other hand, if you ask for an Ethernet card, they will immediately offer you at least half a dozen types and makes to choose from.

The Ethernet card -- in the form of a standard PCI card or a PCMCIA card -- has a port that is known as the RJ-45. Just plug the network cable into this port, and you can connect your PC or notebook to a router, a switch, a printer, a cable modem or, of course, another computer. Incidentally, to connect one computer directly to another, you will need a crossover cable, in which two of the strands need to be intertwined before they are clamped into the pins of the RJ-45 plug.

When was the Ethernet born? It was conceived 30 years ago by Robert Metcalfe, who was working at the famous Xerox Palo Alto Research Center. Xerox PARC was a breeding place for a lot of revolutionary technologies that shaped the way we use our computers today.

On May 22, thirty years ago, Bob wrote a memo to his colleague describing a protocol that would allow computers to exchange packets using carrier sensing, collision detection, randomized retransmission and other technical terms that would interest only students in computer science and electrical engineering. At that time, they were developing the first laser printer, and they needed a way to let multiple computers access the device they were working on.

Over the next three decades, Ethernet had evolved so tremendously that even Bob Metcalfe himself was amazed by it. The Ethernet cards that we find in our PCs and notebooks today are usually the 10/100 version. It means that the card supports two standard speeds: the 10 megabits per second (Mbit/s) and the faster 100 Mbit/s. However, we already have the one-gigabit and even ten-gigabit Ethernet.

The 10/100 Ethernet port has become a standard feature in today's computers. Even bare-bone systems -- computers in mini housing that buyers can configure with the components of their own choice -- come already equipped with it. Even low-cost notebook computers also have Ethernet ports along with a few USB ports. It simply shows the pervasiveness of the technology.

Although I have not been able to build a wireless network to connect the four PCs we have at home, at least the three of us in our family have been able to share one broadband Internet access, thanks to the Ethernet.

And if Gordon Moore came up with Moore's Law, then the now 57- year-old Metcalfe has also been made immortal with Metcalfe's Law. It states that the value of a network grows as the square of the number of the members. Just build your own home network and you will agree with Metcalfe.

* The future of network

The network itself keeps evolving. Not only has the speed increased exponentially, but the intelligence of the devices that manage the network and connect it to a larger network and even to a network of networks has also advanced significantly. These devices are usually called the switches; smart computers whose tasks include making sure that data sent by one computer on the network will reach its destination on the most efficient route and at the highest speed achievable.

Last week, Alex Filloca, the AP marketing manager for Alcatel's eBusiness Networking Division, was very gracious about spending some time with me to explain a few of the current trends in networking technology.

One of these trends is the concept of Quality of Service (QoS). It means that a network has the ability to provide different levels of service in terms of bandwidth, etc., and it has enough intelligence to determine which level should be provided to a particular application.

For example, applications such as e-mail, data backup and virus checking do not really require a lot of bandwidth all the time. On the other hand, applications, such as IP telephony -- which Alex believes to be not very far away now -- video conferencing over IP and multimedia collaboration in product development, do require the most network resources they can use. An intelligent network with QoS capability would be able to identify what applications are active, give priority to the mission-critical applications and, if necessary, take the resources away from the less important applications that can wait.

"If you have a slight delay in your telephone conversation because of bandwidth congestion, you will get distracted," explained Alex, "but if your e-mail arrives 45 seconds late, you will not even notice it."

Virtual Local Area Network (VLAN) is another concept that I learned about during our conversation. In a typical enterprise, a work group or an entire division may have to operate in different locations. However, they usually have only one single physical network -- with Ethernet and perhaps wireless links -- in their organization. A smart network allows us to build one or more of the VLANs that reside inside the physical LAN, but restricts the access of the users who are assigned to one VLAN but not the others.

So, for example, if I worked in the finance division of my organization, I would have the access to the VLAN of this division but I would not be allowed to access the data in the human resources division, which has its own VLAN.

Another concept that I also learned in the area of smart networks is the partitioning. As the IT support team usually consists of people with different levels of skill and knowledge -- and, more importantly, different levels of authority -- the partitioning lets us define what each of them can do. Routine network maintenance, for example, can be left to the junior technical support staff, while network fine-tuning should be left to the most senior staff.

"One of the most important features of the network is the carrier-class," Alex said. It means that the network should be available 99.999 percent of the time as expected by users, he said. It reminds me of the analogy some people have made between the dial tone and the IP tone. Each time we pick up the phone, we expect to hear a dial tone. Similarly, every time we want to access the Internet we expect the network to be available. Of course, those of us who have been using the cable Internet service from Kabelvision have found out that the IP tone is not always available -- even though we have to pay the cable company the same amount every month.

One of the things that should be done to minimize service disruption is to incorporate a "self-healing" feature into the network. A smart network, for example, would be able to detect a misconfiguration or any other kind of human error in setting the configuration. "Research has found out that 70 percent of network downtime is caused by this type of error rather than by the failure of the physical components," Alex said. So, if a misconfiguration is detected, a smart network would be able to automatically roll back and reconfigure itself using the last parameters so that the network can continue to function.

Alex, of course, referred time and time again to his new products -- Alcatel OmniSwitch 6600, 7700 and 8800 -- that boast all of these features plus more. But to those of us who are not enterprise users, which is the target of these products, their features serve to give us a broad picture of how networks have evolved from the day Bob Metcalfe first figured out the Ethernet protocol.