How should an enterprise pick products to build its network? Do they look for the best of each product category, knowing this will increase both integration issues and finger-pointing? Do they select the best vendor overall, knowing that this will invite vendor lock-in and compromises in each product category?\nThis issue is as old as networking, and we\u2019ve still not resolved it. How do enterprises decide when to add a vendor in hopes of getting the best technology, and protect themselves from the consequences?\n\nEvery network vendor wants to be your only vendor. No network vendor wants to accept responsibility for problems, and most don\u2019t even want to work hard to find out who\u2019s causing them. Big network-equipment vendors have not only fallen behind on innovation, they work to actively stifle it, fearing it could damage their incumbent position. These are the views of enterprises, whether they favor single-vendor networks or best-of-breed.\nThe risk of your vendor going out of business is significant unless you buy from a giant. Startups are one-trick ponies that have seized on a niche problem and solved it but will never do anything else useful. No vendors really support standard interfaces or practices, so integrating vendors is an unsolvable problem. The total cost of integration and support for multi-vendor networks will exceed the cost of the hardware. These are also enterprise views, and again from both sides of the debate.\nYou can see from these points why managed network services are getting more popular, but even there, enterprises say that it\u2019s nearly impossible to find a managed service provider that will take responsibility for LAN and WAN, and early reports on those that do suggest that the MSPs themselves are falling into the single-vendor-or-best-of-breed pit, too.\u00a0 We can\u2019t forego networks, so what do we do?\u00a0 Fortunately, the two sides of the debate agree on some points.\nDon\u2019t simply reject single-vendor networking\nThe first point they agree on is that insisting on multiple vendors just to avoid depending on a single vendor is a mistake. You don\u2019t really gain leverage over them, and you multiply your integration and problem-resolution challenges. The only valid reason for a multi-vendor network is that there\u2019s a significant cost, performance, or feature benefit to the decision. Don\u2019t use intangibles to justify \u201cbest-of-breed\u201d network additions. Demand real financial quantification of each potential benefit. The fewer vendors you already have, the greater the savings you should demand. Enterprises are mixed on a specific number, but most say that unless you can show quantifiable financial benefits of between 15% (if you already have multiple vendors) and 30% (if you only have one vendor now), it\u2019s almost certainly not worth adding another vendor or technology.\u00a0\n\u201cAlmost certainly\u201d isn\u2019t a sure thing, of course. Many best-of-breed advocates say that their primary vendors drive them to a multi-vendor approach one little step at a time by employing tactics aimed at locking them in. They say that there are warning signs that you should look for, signs that your vendor is taking you in the wrong direction:\n\nImproved features or performance are available only in new products, new software, and at extra costs. You can\u2019t stay the course.\nYour operations costs are increasing year over year because of increased complexity in the vendor\u2019s management requirements.\nFeatures you need are not available from your vendor until well after they\u2019re offered by other vendors.\nYour problem response times are lengthening, and your problem escalation procedures don\u2019t seem to get your vendor\u2019s attention any longer.\nYour vendor is suddenly reluctant to make any pricing or terms concessions during negotiations for new equipment, or during a refresh of old technology.\n\nIndications that your vendor is taking you in the wrong direction doesn\u2019t mean that you couldn\u2019t make things even worse. Suppose you either see a clear, quantifiable, justification for a best-of-breed vendor addition to your network, or that you\u2019re seeing some of these warning signs. Is it possible to avoid all the risks of multiple vendors and still gain best-of-breed benefits? Probably.\nEnterprises that successfully navigate that tightrope say that the first step is to think enclaves. If you have multiple vendors, keep them in contained areas and connect those enclaves with the simplest and most mature interfaces available. Most networks aren\u2019t vast mixtures of devices creating a single ecosystem, they\u2019re made up of collections of devices performing a common function. One router vendor in core, another at the edge.\u00a0 One data-center switch vendor and another in workgroups or the WAN. Treat each enclave as a discrete network.\nEnterprises already often use different vendors for different enclaves, picking vendors whose cost and features are optimum for each mission. Many vendors focus on specific missions, and that\u2019s what often creates \u201cbest of breed\u201d options in the first place. The interactions of devices within each of these enclaves is typically more complex than the interactions across enclave boundaries, so having a single vendor and even a single class of device within an enclave is more important. Intermingling vendors within an enclave network invites all the classic finger-pointing and integration problems. It will also likely prevent you from realizing any best-of-breed benefits, because special capabilities will be short-stopped by devices that don\u2019t support them.\nAnother useful strategy is to focus on monitoring the interfaces between your enclaves and develop fault-isolation procedures designed to identify which side of the interface isn\u2019t performing. While a single management system across an entire network may seem helpful, it\u2019s often not even useful. Data-center networks are managed differently than branch networks or VPNs. It\u2019s better to focus any special monitoring tools at the enclave boundaries. If you can isolate problems at the boundary points, you can then use specialized management tools to correct the enclave that is broken.\nEven if you don\u2019t have different vendors in different enclaves, this kind of management partitioning often improves operations efficiency. It also keeps the door open for introducing a best-of-breed solution in an enclave with minimal disruption of your operations processes.\nThe final strategy is to accept reality. Despite the debate over single-vendor versus best-of-breed, almost every enterprise ends up with multi-vendor networks, and say that adopting the latter approach in some cases is the right answer. Enterprises don\u2019t want a single-vendor or best-of-breed network; they want an optimum network.\u00a0 The fact that there are so few enterprises that have really settled on a single supplier of network gear says it all. That\u2019s why it\u2019s smart to declare that best-of-breed has won, and focus on how to justify, integrate, and manage the resulting multi-vendor environment. If enterprise experience is a guide, that will get you the optimum network you want.