Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 11 additions & 6 deletions overlay/design.rst
Original file line number Diff line number Diff line change
Expand Up @@ -8,8 +8,8 @@ the new feature to the existing network? There are as many answers as
there are overlays, but they all boil down to one pragmatic
consideration: it is difficult to add new functionality to an
operational network. The challenge of getting everyone to agree to
such an upgrade, let alone dealing with the logistics of deploying the
upgrade across thousand of switches, is an extremely high barrier.
such a change, let alone dealing with the logistics of deploying the
upgrade across thousands of switches, is an extremely high barrier.
Demonstrating the value of a new feature is easier if you can run the
new network *over the top*, without having to modify any of the
existing network infrastructure.
Expand All @@ -26,7 +26,7 @@ in a phenomenon that Clayton Christensen famously called the
*Innovator's Dilemma*. Fortunately, networking gives us a workaround,
with overlays being the widely accepted as a way to introduce
disruptive technology. A National Academies report made this
observation about the state of the Internet over two decades ago:
observation about the Internet over two decades ago:

*The existing core IP network could be used simply as a data
transport service, and disruptive technology could be implemented as
Expand Down Expand Up @@ -83,11 +83,16 @@ is little appetite for augmenting routers with storage. As we will see
in Section |Overlay|.2, however, CDNs do include a "request routing"
function, which has been proposed as a "content-based addressing"
extension to the Internet's forwarding mechanism. But the success of
CDNs in providing the same service as an overlay renders this issue
moot.
CDNs in providing the same service as an overlay has made this a tough
sell. The ease of deploying any new feature in an overlay means you
need a compelling argument for instead adding that feature to the
core.

.. TODO -- some might see this as dismissive of content networking
research, we could consider a sidebar (or not)
research, we could consider a sidebar (or not).

I softened the language, and added a bit more about the bar being
raised. --llp

The second example, video conferencing, supports a *multicast*
function, whereby packets are delivered to multiple end-points instead
Expand Down