Since inception, the resilience of the Internet has been based upon multiple paths from source to destination. At the granularity of the AS, this is generally implemented by linking to more than one AS's. Without resolving this further into whether each link carries advertisements of all or only a subset of routes, this is the basis of resilience.

With particular reference to routers in Tier 1 networks, I am curious about whether the router at the end of each link is implemented as part of a set of redundant routers. I am familiar with VRRP and HSRP for the operation of such a set of redundant routers but I only know of their use in stub ASes, like those owned by end users who need to ensure the resilience of their connection to the Internet.

In other words, how common is the use of device redundancy as a means complementary to multiple links in order to strengthen the resilience of the Internet? I would guess that in cases where links between geographical regions are relatively few (such as links based on submarine cables), device redundancy is an important factor in the resilience of the Internet. This guess is based on the assumption that the portion of inter-region traffic carried by each such link is high, increasing the importance of redundancy at the end point - the router.

  • Redundancy aside I would expect that on the busy links they would have to have multiple routers at each end anyway simply to handle the traffic volume. – Peter Green Nov 2 '15 at 3:19

This would be entirely up to the Tier 1 and their strategy/business model and each one may handle this differently. Redundancy is about preventing loss of service first and preventing degradation of service second (and often a distant second). This means that the focus needs to be on avoiding anything that provides a "single point of failure."

Lets look at your example of the submarine cable. Yes, they could have redundant devices, but it might be more common that they have a single device with redundancy built in (multiple SUP/RP, redundant power) connected to multiple fiber pairs in that cable (probably tied to different line cards/modules in the device). This provides you with most of the redundancy of multiple hardware platforms.

However since the cable itself is a possible point of failure, instead of a second piece of redundant hardware on both ends tied to the same point of failure, it would be better to spend that money putting hardware on a second alternative path (i.e. another submarine cable following a different route). Or to plan for and depend on peering partners that have an alternative path.

While this path may be longer/less efficient/lower bandwidth or less desirable for a number of reasons, it provides an entirely different alternate path should something happen to the cable itself.

| improve this answer | |
  • Your answer is very helpful as it elicits emphasis on the independence of the variables that affect the decision. Whether the Tier 1 operator implements link or device/module redundancy is indeed a multi-dimensional issue. The problem of implementation has at least 2 dimensions: (a)The strategy and business model (b) The depth of resilience required – user3341973 Nov 2 '15 at 11:26

when you think in internet service stability and resilience (as you describe) you need to think in different ISP , not multiple AS , and may you run each ISP on different Router and then run any internal redundancy protocol like HSRP or VRRP between those routers to provide single internet GW to the users.if you run multiple AS on the same ISP it won't make seance cause if this ISP connection is drooped so the all AS related to it will drooped as well.

| improve this answer | |
  • The difficulty may not have been expressed well. The interest lies in what implementation techniques an operator uses to provide resilience. Although I appreciate the difference you emphasise between ISP and AS, that is a refinement of my statement that does not address the core problem. What would be interesting would be to learn whether HSRP and VRRP are operated outside stubs and Tier 3 ISPs. In your description, you seem to be re-enforcing my understanding that these protocols are used by end-users "who need to ensure the resilience of their connection to the Internet" (see OP). – user3341973 Nov 2 '15 at 11:35

Physical device and link separation is key to a redundancy plan. Not only is it beneficial to unexpected failures, it can be critical to uptime when doing planned software upgrades or hardware replacement.

You could have redundant paths to the same IP or leverage BGP multi homing to use different ISPs.

| improve this answer | |
  • I'm interested in redundancy at Tier 1, rather than using different ISPs for end-users "who need to ensure the resilience of their connection to the Internet" (see my OP) – user3341973 Nov 2 '15 at 11:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.