This papers surveys the myriad issues involved in the process of
formation of peering relationships between various ISPs. Instead of
concentrating on the technical issues involved, the author provides a
high level insight into the view point of an ISP on the entire process
of formation of peering relationship. The author starts by explaining
the motivations (reduction in user perceived latencies, cost of
operations etc.) for a new peering. He then touches upon the facts
(traffic asymmetry leading to negligible benefits, loss of transit
revenue etc) which might dissuade an ISP from forming a peering
relationship.

Next the author outlines the major steps in formation of a peering
relationship, viz.,:

 1. Identification of most promising candidates which mainly depends on
the predicted volumes of traffic inter-flow with the peer and associated
gains in revenue (reduction in transit traffic). This apart it also
depends on the ISP's internal policies on peering and broader business
arrangements.

 2. Contacting the right person at the prospective peer ISP and
preliminary negotiation about peering proposal. The negotiation
essentially involves proposing a peering relationship and providing
evidence ( traffic data related statistics)  in support of the proposal.

 3. Deciding on the exact methodology of implementing the peering. This
step involves selection of location, type of interconnection (direct
circuit Vs. Exchange based) and policies regarding operation and
maintenance of the peering. The decisions made at this stage are
influenced by the goal of making the peering work as soon as possible
and to ensure smooth operation over time at the least possible cost.

Takeaways:

1. The decision to form peering with another ISP is a hard one to make
and involves consideration of a host of issues, not all of which are
related to networking.

2. The business goal of an ISP might not always align with the goal
enhancing the overall Internet performance. For example, the latter goal
might dictate formation of peering connections with a rival ISP (leading
to enhanced experience for end users) but this might turn out to be more
beneficial to the rival and hence, against the former goal. This
underscores the importance of devising strategies to ensure that the two
goals are as much aligned as possible, such schemes seem to be missing
in the current state-of-art.

3. The author mentions that intuition to be the main factor on deciding
the usefulness of a peering relationship and the location of the
interconnection. This is because of the huge amount of data processing
involved in carrying out proper statistical studies. However, recent
advancements in algorithms like data streaming algorithms might lead to
more efficient methods to provide better information about the traffic
patterns. This in turn would help facilitate the decision process.