04-10-2019 08:24 PM
What is the difference between receiving full updates and a default route with BGP? All networks should be reachable with a default route to the ISP. What is the benefit of receiving full updates compared to having a default route out to the ISP?
04-10-2019 08:41 PM
Ddefault route - is the route that will be used if there's no other route that matches the destination in the router's forwarding table - in this case all go to ISP and from there it will go different path based on the ISP routing decision.
If you have single ISP this is more than Enough
Full routing table - is a table which contain all the routes the BGP neighbor is aware of from your ISP's
If you have Multiple ISP you peer with, then with the Full routing table, you can do Traffic Engineering.
Required More resource on Router for the Full routing Table.
04-10-2019 09:40 PM
If you have only a single link - there's no difference, so just get a default route.
If you have dual links then some sites should be quicker to get to via ISP1, and some quicker via ISP2, and getting a "full" table lets the router learn this. The downside is you need a lot more memory to hold to full tables.
04-10-2019 10:25 PM
Hi,
A default route is a route that will be used if there's no other route that matches the destination in the router's forwarding table.
A full routing table is a table which contains all the routes the BGP neighbor is aware of.
If you are connected to the Internet through a single ISP, a default route is enough, since there's only one possible path, so there's no point in having 771056 routes in memory that all point to the same next-hop.
If you are connected through several ISPs with BGP and want to always use the "best" path, then a full routing table makes sense. In this case, the default route will (almost) never be used since the router knows every possible destination with a specific route.
However, a full routing table takes some memory and also more CPU power to perform a lookup for each destination among the 771056 known routes.
04-11-2019 12:55 AM
Hello
just like to add to the other comments that even if you receive defaults from multiple isps it doesn’t mean to say you cannot manipulate your egress traffic - you could possibly still have an option through the use of policy based routing.
04-11-2019 05:06 AM
04-11-2019 05:08 AM - edited 04-11-2019 05:13 AM
The differences lie in trade-offs.
A full internet table (+700K routes) will give you granularity from a traffic engineering standpoint. However, if you only have one provider (i.e. single-homed), there is no functional difference, your egress point will be the same.
The situation changes, however, if you have contracted two providers (i.e. dual-homed). Each provider would have a different view of the routes, and, by receiving the full internet table from one of them or from both, you can achieve efficient outbound and inbound policies. That is, having complete control of how your traffic leaves your administrative domain, and being able to influence (note the difference between complete control and influence) how the traffic enters your administrative domain.
In your case, your administrative domain is your AS, you can always decide how to route the traffic out of your AS, but, as with every fair game, the same rules apply to your provider(s). That's the reason because you can only influence, but not control the inbound traffic flows, they have the last word as the traffic is leaving their AS.
Now, how this applies to the full table, partial table or only default? The granularity in BGP policies can go down to a per-prefix policy (i.e. subnet/network) through attribute manipulation. Having a full routing table, or even partial (usually regional services) table allows you to have control of this policy, you are free to implement it and be as granular as you'd like. The downside is the amount of resources you dedicate to this. Approximately, 100K prefixes occupy 70MB in your memory. Doing the math: at least half GB per full routing table, assuming you'd have 2 copies (one per provider), then boils down to 1GB ONLY to keep the routes, nothing mentioned about policies or attributes. Please note that these numbers are approximations, take them with a grain of salt. Also, remember the previous statement is only about BGP, you would have many more features/protocols running in your router, all of them will consume space in memory to run.
Ultimately, this will lead undeniably, to resource consumption. The question is: are you in the position to face this trade-off? BGP is a very heavy protocol to run, and granularity is achieved thanks to resource allocation :)
Its totally up to you and your current needs.
As a personal opinion, having only one provider only and lacking the luxury of high resources in the router, doesnt make it worth the effort.
Hope that helps,
David.
05-18-2023 05:42 PM
I can see all the replies and it seems that from my understanding, traffic engineering with more granularity means having the capability to control which AS you want to receive or advertise your network to, is that right?
I believe you can still do that with AS prepending (incoming traffic) and LP for outgoing traffic even if you have 2 ISPs and both are just advertising default routes to your BGP router. Please let me know if this is also a feasible option for 2 ISPs with BGP route manipulation as a requirement.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide