[Ila] BOF proposal draft

[Tom Herbert <[e-mail]>]

Hello,

Here is dome draft text for a BOF proposal description. Please comment!

Thanks,
Tom
-----------------------------
The goal of this group is to standardize Identifier-Locator Addressing (ILA).

The problem to be addressed by this group is how to provide network
overlays with high efficiency, low latency, scaling to billions of
users, seamless mobility, strong privacy and security guarantees, be
interoperable with existing networks, be anchorless, be usable for a
variety of use cases, and have simplified control and management.
These requirements are being driven by a huge growth in number of
connected devices, particularly IoT, as well as the emergence of next
generation applications such as AR and VR that have stringent
networking requirements. While there have been many solutions proposed
and defined by IETF to provide network overlays, including countless
variations of encapsulation, NAT, and segment routing-- we believe
none of these have been shown meet all these requirements.

ILA is a type of identifier/locator split that partitions an IPv6
address into two components. One part of an address expresses the
immutable identity of a node, and another part indicates the
topological location of a node which can be dynamic. ILA is a means to
implement network overlays without the overhead or complexity of
encapsulation or extension headers. ILA leverages the immense size of
the IPv6 address space.  In this regard, ILA bears similarity to
Identifier-Locator Network Protocol (ILNP), however there are some
major differences in protocol layer and the control plane.

There are three primary use cases proposed for Identifier Locator Addressing:

  * Mobile user-plane

  * Datacenter virtualization

  * Network virtualization

A recent trend in the industry is to build converged networks that
contain all three of these use cases. In order to meet very low
latency requirements and provide high availability, many mobile
providers are housing data centers within their networks to run
applications. Similarly, many providers also run cloud services within
their network. A single networking solution with a common control
plane is desirable.

There are two aspects to ILA: a data plane and control plane. The data
plane includes the ILA address transformation mechanism, checksum
neutral handling, and ancillary protocol support (such as
considerations of ICMP when ILA addresses are involved). A
transformation is drive by a mapping lookup on an identifier to a
locator. The control plane’s main focus is on the mapping system.
There are two critical problems to be solved in a mapping system 1)
scaling to support billions of entries 2) securing information in the
mapping system which is inherently privacy sensitive data.  The
mapping system is essentially a key-value store database. There has
been a lot of innovation in key-value store solutions over the past
few years,  one goal of this group will be to evaluate key-value
solutions for providing the mapping system.

This activity may leverage protocols and work from working groups
within IETF including dmm, nvo3, lisp, 6man, and int-area.
Additionally, work in other SDOs, particularly 3GPP, is relevant. Open
source projects, particular work in key-value store technology and
databases, are also relevant.