
By Chris Lewis
Even corporations that profited smartly from the U.S. economy's recent bull run understand that a similar growth rate may not occur in the coming decade. Savvy organizations are studying ways to sustain corporate growth, and expansion into international waters is a popular option. Internationalization is in vogue, driven simultaneously by deregulation within Europe and the potential of Latin American markets. Given that modern telecommunications companies exist in virtually every developed country, you would expect that expanding a U.S.-based network into other countries to support this international drive would be fairly straightforward, right? You couldn't be more wrong.
Despite the multitude of companies that claim to offer full international reach, there still is no one-stop shop that extends a network of its own cables to all the international destinations you'll need. Global telecom companies, such as WorldCom or Global One, support only a limited number of cable-serviced locations. To reach many international locations, they must sign agreements with other carriers, which dilutes the benefits they can provide in comparison with bandwidth resellers like GAINS International or IXnet. A chief advantage of companies like WorldCom is that they own and operate their own network cables, so you know they're in control of the levels of service they provide. Resellers, on the other hand, don't own any cables; they lease bandwidth from carriers and pass volume discounts onto clients, and therefore depend on resources out of their control to provide service. However, if one of the telecom companies must use a third party to reach some of your destinations, it is behaving much the same as a bandwidth reseller.
Here are some key issues underlying an international expansion and what to expect from vendors:
Key Issues To Consider
In our article on designing fault-tolerant IP networks (www.networkcomputing.com/806/806ws2.html), we analyze what it takes to craft backbone and distribution networks, such as network design and interface standards. For the most part, the same issues apply here, with the most significant difference being the cost differential for international bandwidth. Within the United States, it's feasible for large networks to rely on a T-1 backbone that provides flexibility when dealing with fault scenarios. Spare capacity on backbone links during normal operation lets rerouted traffic be carried without congestion.
By contrast, the cost of international T-1 class links is astronomical. Traversing the Atlantic to London isn't bad, but as soon as you branch out into the mainland, be prepared for serious sticker shock. To illustrate this, GAINS International prepared a proposal for an international WAN deployment that is competitively priced given the current market conditions. Even so, GAINS indicates that it costs $28,000 per month for two diversely routed 1,024-Kbps links from New York to London. In the same proposal, the cost for a single 256-Kbps link from London to Madrid is $9,900, and a single 256-Kbps link from London to Tel Aviv costs $14,900 per month.
So, besides cost, what are the issues to consider? Obviously, the first is network design. If you don't want to design your backbone, you might consider companies that offer international frame relay solutions. However, if you've ever been disappointed by domestic frame relay services' failure to deliver anticipated burst rates or throughput, you'll find it difficult to deal with international frame relay. These days, if you really need 128 Kbps of throughput, you need to attain a CIR (committed information rate) near that figure, which can be more expensive than clear bandwidth at that level.
If you opt to design your own backbone, you'll need to weigh a whole host of issues. Assuming you craft your backbone links to carry rerouted traffic (in the event of link failures) as well as usual traffic, the link will exceed 64 Kbps, which calls for E-1, not T-1, services. E-1s are a lot like T-1s and come in channelized and fractional varieties. The main difference is that an E-1 has 2.048 Mbps of throughput, which is half a megabit more than the U.S. T-1. The E-1 is still split into 64-Kbps DSO segments, 32 of them to be exact. However, two are reserved, leaving you with up to 30 available channels.
PRI services are also delivered over E-1s, but here things in the international arena start to get messy. In Europe, there are many kinds of standards for termination and signaling. An example of a country-specific standard that can cause problems is British Telecom's DASS-2.
DASS-2 is a signaling standard developed by Plessey and used by British Telecom to deliver PRI services to PBXes for use in voice networks; it does not conform to Q.932 for ISDN. Router equipment manufacturers do not support that standard, and typically require I.421 in Europe, TS014 in Australia and 1TR21 in Germany. This is particularly tricky because PRI can be delivered using G.703 framing with a coaxial cable, or the more familiar RJ-45 presentation for I.421. In Britain, for example, you order PRI for PBXes with G.703 framing and coaxial termination, but PRI for data should be ordered with I.421 and RJ-45 termination. It's easy to order the wrong service if you're unfamiliar with all the options in these locations.
|