[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RRG] New RG charter discussion
Hi Dimitri,
the sentence
"At the moment, the Internet routing and addressing architecture is
facing
challenges in scalability, mobility, multi-homing, and traffic
engineering."
is ambiguous, that statement tends to bundle causes and
consequences and
imho creates an amalgam in the problem statement the RRG is going
to work
on:
Agreed. This is a necessity, as the charter is by nature a broad and
inclusive statement.
o) routing system problems
- stationary: scalability (RIB & FIB size)
- transient: efficiency updates (add, remove) frequency, and
resulting convergence time
o) root architectural causes
- protocol dual stack (due to address space scaling)
- address = locator (result in traffic engineering issue ~ meshedness)
use address as identifier (result in policy routing issue)
- mobility and multicast (overlay applications)
o) operational causes
- traffic engineering leading to prefix de-aggregation
- policy routing leading to prefix de-aggregation
- dual/multi homing (for resiliency)
- address allocation (due to address allocation policy)
but the major problem is not there:
o) FIB capacity combined to improved search algorithm solves the
stationary problem not transient problem (i.e. routing dynamics), the
latter is where IETF can play a role, the former is a vendor-specific
system engineering design - hence the question becomes where/what
is the
role of the research community ?
The architectural deficiencies that you cite above become the drivers
for FIB growth that cannot be economically sustained. Thus, the
thought is that by producing a more scalable architecture, we can
reduce the FIB growth rate.
o) if system engineering can come to a solution for the stationary
problem
(without complete routing engine re-design) it still does not mean it
solves the causes, i.e, such solution is not going to eliminate the
need
to support for mobility, multicast or traffic-engineering (hence the
latter become part of the constraints to be taken into account)
Agreed.
o) one can also notice that the slope of the curve of increasing
number of
prefixes is also important to analyse (growth rate) vs system
capabilities/needs - for the time being, the internet community at
large
can only benefit from ad-hoc measurement or unreliable predictions
(based
on rough extrapolations) in order to derive such behaviour - from that
perspective, it is noticeable that no effort is considered to
provide for
a reliable and robust (i.e. scientific) methodology to analyze the
phenomenon itself - this means that as chartered, the RRG will
never be
able to play a real role in predicting / anticipating such
"consequences"
but also RRG will never be able to confirm or infirm engineering
community
expectations
I think that the RRG is a fine venue for collaboration, but the
ongoing study and monitoring of the DFZ table is a non-trivial
undertaking that will require funding. As the RRG is not a funding
organization, it can at most connect the funding with the researchers
to put this type of study in place. This is certainly something that
I'd like to see happen.
hence, before focusing on engineering long or short term solutions
(role
of the IETF), the IRTF shall focus on real research tasks, such tasks
start by asking what are the causes of the problem and how the current
system with its goodies and flaws can be modeled - in brief, the
role of
the IRTF is to ask WHY such behaviour occurs, identify causes of the
problem and model global system behaviour
I have no problems with further investigation into the details of the
problem. However, some of us are already quite convinced (through
first hand observation) that there is a real, tangible problem that
needs to be addressed. As a result, I'm not willing to completely
serialize our work dependent on the outcome of research into the
problem. Thus, we should simply push on in parallel.
the lack of systematic tools is a REAL issue that the research
community
at large shall address before trying to work on engineering oriented
solutions
I agree that more tools would be welcome.
Tony
--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg