How Autonomous Internet enables 100% uptime
A big advantage of Autonomous Internet, in comparison to traditional provider-based Internet, is that it makes 100% uptime of your network possible via redundant Autonomous Routing technology. The main features that enable such high resilience are described here.
Multiple connections to the Internet
Multiple physical paths/links
One can have multiple, fully resilient, simultaneously active route paths to the Internet via separate physical links, with sub-millisecond failover. So if there is a problem reaching a destination via one path/link, there is always a second path available on a different link to route the traffic over. Technically this is implemented by 'BGP' (Border Gateway Protocol), and specifically eBGP with an 'e' for BGP sessions with external parties.
Multiple (transit) carriers
By using multiple different transit carriers (to reach all National and International Internet destinations), a problem that is contained within the network of one carrier can simply be circumvented by using a different second carrier. A second carrier has a different bacbone, a different network architecture and layout, potentially different router/switch hardware and software, different maintenance windows, a different organization with different people, etc. Therefor using a second link from a different carrier is much more useful then using two links from the same carrier.
Multiple BGP Routers as gateway
A network can use multiple (BGP) routers instead of one. In case one of the routers fails or is under maintenance, all Internet traffic can seamlessly continue flowing via the other router.
Externally, both routers have at least one transit connection (for global/international traffic), so each of them can reach the entire world independent of the other router.
During normal operation both routers exchange their eBGP routing information with each other via iBGP (with an 'i' for an internal BGP session), so both of them can still use all paths available via the other router as if they where local to that router.
During an outage of one of the routers, all traffic is routed over the links that are directly connected to the other router, making sure connectivity stays up.
VRRP for Default Gateway redundancy
When using two BGP routers, internal servers can configure either router as their 'default gateway'. Both routers back up each others internal IP address using 'VRRP' (Virtual Router Redundancy Protocol) or other redundancy techniques. If either router becomes unavailable, the other router takes over its IP address within 3 seconds, making sure internal traffic is still routed on the Internet.
Advanced redundant performance options
Instead of using one link as primary, and the other only as hot standby backup, it is also possible to use multiple links in a load balancing configuration. In that case for each individual destination on the Internet the best and shortest available path is selected, maximizing performance.
The actual distribution of traffic over the links is defined by the relative quality of the paths on each link: it is not a router setting. So it is not possible to set exact percentages of how much traffic each link should get. However, with traffic engineering it is possible to influence the load distribution by changing the quality parameters of paths.
Improve routing via direct peerings
By using not only transit links but also setting up direct peerings with other local parties on an Internet Exchange (IX) it is possible to further improve the routing quality and resilience.
Via peering intermediate transit carriers of either party are removed from the path between the peers, and inherently cannot negatively influence the performance or availability of the routing between the peers anymore.
There is no shorter and thus more reliable way to reach other parties then peering: direct peering always results in shorter paths and less equipment in between which can break.
Peering via partial transit
Setting up direct one-on-one peering with a substantial amount of other parties is a lot of work and maintenance, and ability to do so depends on the terms and conditions other parties set for peering (e.g. minimum amount of traffic volume).
An alternative for setting up your own direct peerings is using partial transit. In that case a specialized carrier offers a service where routing paths are provided, not to all global Internet destinations as usual, but specifically to only local destinations. For example an "NL Transit" service can contain routing to all Dutch destinations, only via direct peerings of that carrier.
With a partial transit service you have most of the quality and resilience advantages of peering, without the related workload and maintenance of doing it yourself. Furthermore it will probably include access to peerings you would otherwise not be able to get directly due to the terms and conditions of some peers.
Cross service backup
Backing up peering via transit
With transit you can reach all Internet destinations, both International and National. Therefor, transit is automatically a backup for peering and/or partial transit, albeit with higher cost and possibly degraded performance.
Setting up a configuration with dual/redundant direct peering or partial transit links for resilience purposes is therefor only needed in a situation where the traffic volume and capacity of the National (peering, partial transit) links is substantially larger then the capacity of the International (transit) links installed. E.g. 80% National traffic and 20% International traffic. In that case transit simply does not have the capacity and performance to be an adequate backup for peering or partial transit.
The above arguments clearly illustrate the substantial technical advantages and Non-Stop character Autonomous Internet offers.
For more information please contact us: we are happy to talk the whole process through in advance.