Technical Intro

beginner intro advanced intro operation

Beginner Intro

The Internet consists of more than 10,000 separate networks, each managed by a different team following their own implementation policies.  Each network is connected to one or more others and each such connection is a peering arrangement.  Consequently, the Internet consists of many hundreds of thousands of peering arrangements between the 10,000+ separate networks.

Each peering connection is a two party conversation between a pair routers and each pair exchanges details of the IP addresses they each have.  When setting up a peering connection, there are choices to make.  For example, what do you announce, and what do you accept?  These choices form your routing policy and effect how your customers reach other networks and visa-versa.

History has shown the wider networking community that human error & inexperience can cause global routing problems and this risk is addressed by applying carefully crafted filters at each end of every connection to constrain what is announced and what is accepted.

These filters, being static, do not adapt to ongoing changes in global network topology, and so they have to be maintained.

To help with the process of managing & maintaining all these different static filters, a number of databases have been set up to hold high-level topology information about which network has what, and is connected to who.  Apart from the 'what' and the 'who', these databases also provide a list of customer networks for each network, and this is used as a tree structure detailing a whole raft of other networks which should be visible over a single new peering connection.

However, like most peering policies, these databases are maintained, manually, by each their individual management teams.  Consequently, we have found many of them to contain inaccuracies to a magnitude which undermine their overall value.

With the advent of our Autonomous System Peering Analysis tool, we are periodically collecting real routing information from multiple points around the Internet.  This dataset includes a definitive list of IP address blocks being announced by each network and a view of which networks are seen through which other networks.


Advanced Intro

Supporting and maintaining an ISP's peering policy is a continuous task.  Whether you are highly skilled or early on the BGP learning curve you will likely have spent many hours or days fixing 'simple' peering or filter problems that result from an up-stream's or down-stream's mistakes.

You are busy, your peers are busy, the Internet is growing at an ever increasing rate and you rarely get the opportunity to review the inside and outside views of your network.  We are working to provide you and ISPs like you with various tools to simplify this important task and make it easy to review, regularly, routing leaks and identify errors made by your customers, free peers & transit providers.

With frequent requests for filter updates coming from peers, how sure are you that all your filters are up-to-date?  Do you have the time to deal with them all, as they come in, fully verifying each one?  If you are using AS-MACRO expansion, how confident are you that they represent reality?

Skeptical?  This means you probably understand the complexities involved, so why not take this opportunity to see if these tools prove your peering is the way you want it and check your filters are correct for each of your peers.


Okay, so how does it work ?

First
we regularly gather routing information from viewing points around the Internet and merge it in to a single homogenous dataset for use in the subsequent steps.
Second
we calculate a 'connectivity index' for every visible AS giving us a connectivity ranking for all networks.
Third
we produce reports for any given AS detailing visible address space & apparent transits & customers, noting their calculated rankings.
Fourth
we produce elements of BGP prefix, as-path & access-lists which include all currently advertised address space for any AS and apparent customer networks with worse connectivity rankings.

Customers with better rankings are usually other transits or free-peers but, optionally, these can be included.  However, previous routing mistakes frequently result in the whole Internet being seen behind such customers.

Further information about the AS Analysis Tool can be found here
and
Further information about the BGP Toolkit can be found here.