So we're now going to turn our attention to multi-homing. In this series so far we've learned all about BGP, how it works, and the BGP attributes and how to configure them. What the multi-homing series will do is how to use the BGP features and the BGP attributes to allow our network to connect to other networks. Whether it's one, two, three, four or more other autonomous systems. First off we're going to look at why we multi-home. And after all, maybe our upstream provider will come to us and say, "Just connect to us. We're perfect. 100% reliable. It'll always be there, don't you worry, we will look after you." Have you heard those stories before? But in fact, today multi-homing is very, very important for any operation that's serious about its connectivity and continuity on the internet. Redundancy is one of the important features. One connection to the internet means the network is dependent on the local router. So whether that's the configuration, the software, or the hardware. Configuration is influenced by the operators of the network. Who changed something? Did they change something? What was changed? Or maybe it wasn't changed. What about the software? Is software 100% reliable? 100% perfect? Never fails, never has an issue. So a software outage or a problem caused by the software on the local router means an outage for the local network. What about the hardware? Is the hardware 100% reliable, 100% perfect? And most network operators with experience of using the different variety of network equipment knows that configuration and software and the hardware can all have potential issues over the lifetime of the router. So we don't want to be dependent on the one single router that we have to connect to our service provider. What about the WAN media? This is the link between your network and the upstream provider. It could be a physical failure, could be a carrier failure. Physical failures are entirely possible, whether it's somebody cutting the fiber-optic cable, something goes wrong with the copper, cables get dug up, cables get cut up, cables get stolen. People may believe the fiber-optic is copper and think it has value, and when they discover it is a piece of glass they leave it cut up, the damage has been done. Could be the carrier failure, if you're dependent on upstream provider's layer 2 infrastructure. If that breaks, you have no connectivity to the Internet either and there's probably not much you can do about it until the carrier has fixed the infrastructure. If your internet connection goes over international links, whether by satellite or more commonly today through submarine fiber, we all have heard of the many issues and outages that can affect both satellite communications and the submarine fibers. So redundancy in that respect is considered really, really important by many operators. We've heard the stories of whole countries being cut off simply by a ship's anchor being dragged through the submarine fiber and breaking it. If we've no redundancy, then we're entirely dependent on an upstream service provider as well. And we can go through the same list as we had with a local router. It could be the configuration of their router, that's hosting your link. It could be the software on it, it could be the hardware. Any issue that the upstream service provider has will impact your network as well. And not just the configuration or the software or the hardware, it could be the operation of the upstream service provider. If you're a network operator for example, in Europe, and your sole connection to the Internet is through a network operator in North America, well the North American network operator would do the planned maintenance 4:00 a.m. to 7:00 a.m. quite typically, seems a reasonable time. But that would be U.S. East Coast time, five or six hours time difference from a network operator in Europe. 4:00 a.m. to 7:00 a.m. East Coast U.S. time, would mean 10:00 to 1:00 p.m., yes 10:00 a.m. to 1:00 p.m. in Europe, the height of the business day. And really the worst time for any network operator to experience an outage. So being dependent entirely on your upstream service provider is really not too sensible for many network operators hosting business critical applications. Apart from redundancy we also need to consider reliability. Business critical applications demand continuous availability. If there's no redundancy, it means you don't really have much reliability, and when you don't have much reliability, it means significant loss of revenue. Most network operators today offer some level of availability, 99.<whatever>% is quite common. Back in my early internet days, we were offering 99.5% availability, and that's only a few hours of downtime per year. Today it's quite common to see operators offering 99.9% availability, and that's down to a few minutes per year. If the operator is spending all their revenue refunding customers because they haven't met their availability guarantees, what does it mean for the business? So reliability is of critical importance to most network operations today. And there's also supplier diversity, many businesses demand supplier diversity as a matter of course. Now this can mean internet connection from two or more suppliers, with two or more diverse paths, two or more exit points. And for international connections, two or more international connections, in fact two of everything. And you see a common theme through this whole presentation series. We're talking a lot about two of everything. For multi-homing two of everything is very, very important. Because if one fails for whatever reason, temporarily or longer-term, you have a backup, which means that your network infrastructure and your end-users can carry on using the network even with the disruption in place. Another reason for multi-homing that not many people think about, is for changing upstream provider. Because if you've got one upstream provider and you want to move to another one, it's actually quite involved process. You have to disconnect your existing connection, move the link to the new upstream provider, reconnect the link, reannounce the address space, and hope that this will all work. And how long will this take? Matter of seconds? Unlikely. Minutes? Hours? Days? Well it's hard to tell, but it's certainly not going to be a seamless switchover that your end users are not going to see. And in fact, in my experience a lot of these changes in upstreams can take days, if not weeks to achieve. However, with two upstream providers migration becomes very simple. The operator brings up the link with the new provider, including BGP and all the address space announcements. The operator can make sure that all the filters are working around internet, that the new link will take traffic, do all the integrity testing on the link and so on. And then once they're satisfied with the new link to the new upstream, they can disconnect link with the original upstream provider they're wishing to discontinue service. And this means there's no break in service for the end users, And the network operator can transition from one upstream provider to the other upstream provider without any break in service. I mean this is not really a reason, but often quoted by many of the folks I have done BGP workshops and assistance with network infrastructure for, and that's leverage. In other words, playing one network operator often against another one. For things like service quality, service offerings, and availability. If your upstream provider knows that you're only going to connect to them, do you think they're going to make best effort to give you the absolute best service, and 100% attention all of the time, every time? Unlikely, what's the incentive for them? But if they know that you're connecting to another upstream provider, they know that you're going to be comparing their service with their competition all the time. And so getting two upstream providers is actually a very astute move for any network operator. Really to ensure that they're getting the best price, best service, best connectivity, and best availability. So it's actually quite easy to demand multi-homing as a requirement of any operation. A C-level director, or member of the company, can make this a requirement. But what does it actually mean in real life? What does it mean for the network in question? What does it mean for the Internet? And how do we go about doing this? There's no command that you can insert on the router that says "BGP multi-home". We actually have to sit down, study the BGP attributes, how to implement policies. The relationships between the autonomous systems out on the Internet to work out what is best for our network and how we implement multi-homing.
© Produced by Philip Smith and the Network Startup Resource Center, through the University of Oregon.
Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)
This is a human-readable summary of (and not a substitute for) the license. Disclaimer. You are free to: Share — copy and redistribute the material in any medium or format Adapt — remix, transform, and build upon the material The licensor cannot revoke these freedoms as long as you follow the license terms. Under the following terms: Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. NonCommercial — You may not use the material for commercial purposes. No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.