by John Wobus  Upgrading Your Network Backbone Evaluating proposals Buying equipment based upon future upgrades Vendors often announce features a year or more in advance, and sometimes come out with products that don't yet have all the announced capabilities, leaving them to be added later. While such products have some attraction, the companies may not manage to follow through. Preannounced features are dropped, delayed (sometimes for years) or replaced by better , less costly solutions. Betting that any company will follow through is a significant risk, particularly when you need these future features by a certain time. Bleeding edge versus established technology, features, and protocols When a new technology is introduced, people feel that they can save resources by buying into it early, thereby taking advantage of its whole life cycle rather than shifting into it later. While the ad vantages are self-evident, the potential problems are often over-looked: -It's the memory of successful technology life cycles that suggests this strategy-and only the successes are remembered. Every technology starts out as a dark horse. It is not infrequent for a successful technology to stay that way for some time, while others that seem inevitable are revealed to be faulty or fail. For example, Ethernet 10BASE-T and TCP/IP are technologies that many sites wish they'd bought into earlier, but during the introduction of both of these technologies, there were alternatives that looked just as attractive. -All the successful technologies and their implementations have been improved through the experiences of the early suppliers and adopters. Economies of scale and the cost of converting from one technology to the next dictate that each technology be significantly faster than its predecessor. Product design in the new technology has to accommodate its higher speeds as well as other new factors. However, in reality, a supplier's first priority is getting the first generation out the door, often without even the intention of harnessing any of the additional speed or other new factors the technology is capable of providing. In the second generation, vendors can apply the lessons learned from the first, while having had more time to address issues previously passed over. Subsequent generations will be of higher quality, offer better performance and cost less. People often underestimate the degree to which such early generation hardware limits the performance promised by the technology. In addition to this, the technological standards often are still in evolution during the beginning of the life cycle. In many cases, it is clear that there will be not one, but a series of software or firmware upgrades involved as the standards mature. Dealing with software and firmware upgrades is one of the largest sources of downtime and reliability problems. The strategic decision to adopt a new technology early can inf luence the maintenance load and the reliability of the system for years to come. New model equipment The age of the models of equipment to be purchased brings up the same issues as the age of the technologies: a vendor's very newest models often have compelling new features, capabilities and price-performance, perhaps those that directly address the goals of your upgrade. But new-model equipment often has bugs to be worked out. You could be setting yourself up for months of unreliability and troubleshooting. Site visits: talking to counterparts at other sites Industry labs often test equipment in settings unlike yours, and vendors naturally portray their own products in the best light. So, discussions with and visits to sites similar to your own can be of great value. Even so, it is perfectly possible to visit a site, listen to them rave about a product, then find yourself disappointed in it. You must take into account that different sites tolerate a different amount of support and problems: The site you talk to might think nothing of weekly emergency downtime, or may be supporting their system with three times your staff. Also, sites have a natural tendency to want to portray their own vendor choices as good ones. Another danger is that the site is not using the features that are steering you toward a particular solution. You need to figure out what new features might fit this category so you can ask about them specifically. Trade journal headlines can lead you to believe some particular feature is in common use long before that is the case, so it is better to ask about seemingly pedestrian issues rather than be surprised after the fact. Be sure to ask about a product's ability to compensate for failures. In other words, when a piece of equipment fails (or is removed), how does the system works around it. Such systems often don't work as well as vendors lead you to believe. One thing you can do during a site visit is to ask them to demonstrate the feature on their production system by unplugging something. They won't do it, but they will explain all the reasons why they decline: They will tell you about many of the "gotchas" that it is good to know before the purchase rather than after the fact. Future software/firmware upgrades Often the equipment you buy will have software and/or firmware upgrades after you purchase it. In large-scale networks, this seems almost unavoidable, and the honest network manager will observe that many of the outages that effect most people stem from problems introduced by firmware and software changes. In this area, the simplest scheme without VLANs, ATM, or routing has the advantage: Its firmware is simplest and least likely to require upgrades. When upgrades are necessary, they are solely to fix bugs rather than to introduce function. Unfortunately, in any major network, you are unlikely to use such ideal equipment: The vendor is likely to introduce features even if you don't plan to use them, and will eventually require upgrades if they are to provide support. Future backbone upgrades and expected life With a major backbone equipment replacement, naturally you wish to leave yourself in a position needing only minor upgrades when you need to connect more LANs to the backbone. It is very difficult to predict the life expectancy of equipment. And a surprising number of major upgrades are done in computing and networking which avoid the question all together. It is unlikely that we can accurately predict beyond about three years. While we often aim for a longer lifetime, each additional year we aim for adds quite a bit more uncertainty. Ironically, we are often using equipment longer than we expected (seven to 10 years, perhaps more), sometimes because the need for newer equipment does not arise, and sometimes because of cost factors. Ideally, you can consider any time the equipment continues to fulfill its function beyond three years to be a happy bonus. Automatic Fal lback and Redundancy Over time, the network grows more important: More of the staff depends upon the network for daily work and productivity suffers when network services are not available. Network managers feel compelled to reduce down time and to reduce its risk. Fortunately, like most electronic equipment, many kinds of networking equipment have grown more reliable . One approach to reducing down time is to build redundancy into the backbone. Design it so that if one part fails, there are routes to bypass it. Note that falling back is one of the more difficult things to achieve, and that if you are paying for it, you had best know how well it works. For example, does it work quickly enough to avoid clients from being signed off of servers through some sort of keep-alive mechanisms? Are there jobs that would have to be restarted despite the fallback measures? This is definitely an area where you are more interested in borrowing on someone's experience than in what the vendors tell you. Next Upgrading Your Network Backbone Is It Time To Upgrade? What Do You Need To Know? What Sort Of Upgrade? Additional Issues Updated March 14, 1997 Print This Page E-mail this URL