Techweb
Rash's Judgment
Why You Should Start Thinking About DWDM Technology
By WAYNE RASH
September 15, 2000
If the number of press releases and calls I get are any indication, you're going to be hearing a lot about a technology called DWDM this year. In fact, dense wavelength division multiplexing may be the Next Hot Thing for the enterprise.
DWDM is important because it enables a single optical fiber to carry many times the amount of network traffic it could otherwise. The process works by creating signals from several lasers that emit slightly different wavelengths of light, and combining those so that they travel on one fiber. At the other end of the fiber, the different wavelengths are separated, and the signals recovered.
It sounds simple, but it's a difficult, complex process that until now was available only to carriers.
Now that component manufacturers have created small, relatively inexpensive devices that can multiplex light, this technology is ready for the enterprise.
The most immediate enterprise need for DWDM is for connections between Gigabit Ethernet switches in data centers and ISPs, where the traffic demands continue to grow and the solutions are otherwise limited.
In fact, until now the only way to increase the bandwidth between switches was to add fiber connections. You might find, for example, that a 12-port switch would have four of its ports dedicated to carrying traffic to another 12-port switch.
While this setup works, it soaks up ports and is very expensive. The expense comes from the fact that those ports used to create the high-bandwidth trunk could otherwise be used for other purposes, and because additional fiber runs must be installed and managed.
When you multiplex the signals, you need to provide only a single fiber pair for the high-bandwidth backbone connection. Depending on the distance of the run and other circumstances, this could be significant.
For example, if your network is bandwidth-constrained and you have to pull more fiber to add bandwidth, you could find yourself spending tens of thousands of dollars on labor and other factors just for installation. If you have to go across the street, the figure can jump by an order of magnitude.
On the other hand, if you can increase the amount of traffic your existing fiber carries, you can avoid those costs, as well as their associated delays, and simply plug in a new switch (or maybe just a new blade for your existing switch).
As you might imagine, there are other considerations.
Only a few companies will be shipping this capability in the near term. Among the first is Extreme Networks, which began selling a 10-gigabit multiplexed intra-switch link for its high-end products last spring.
The lack of broad support for DWDM could mean that the manufacturer of your current network infrastructure isn't going to have such products for a while. That could mean you'll have to make some hard choices--you can wait until something becomes available, change to a different network infrastructure (a very expensive proposition) or resign yourself to having multiple infrastructure manufacturers on your network (a work-intensive solution at best).
Adding to the complexity is the fact that DWDM products are sure to be proprietary at first, and they'll also be expensive. Whether meeting your backbone needs can be done more practicably by choosing multiplexing vs. simply adding fiber will depend on the switches and other details of your infrastructure.
Sure, DWDM-capable switches will be expensive at first, but they could very well be cheaper and more attainable than whatever else you might do. When's the last time you found something in networking that was cheaper in any way at all? If your backbone traffic is projected to exceed the capacity of your core anytime soon, perhaps this is the time to start the process of upgrading.
Over the long run, it's a sure bet that you'll move to DWDM or a related technology. You simply can't add fiber forever, if only because you'll run out of places to put it.
Wayne Rash is managing editor/technology. He can be reached atwrash@cmp.com orwrash@mindspring.com |
