2008.02.20 NANOG 42 Lightning talk notes, round 3

Matthew Petach mpetach at netflight.com
Wed Feb 20 18:39:27 UTC 2008


I went ahead and grouped the notes from all
three lightning talks together into one set of
notes, as the speakers were powering through
so quickly there was no time to even save one
file and start a new one.  ^_^;  Again, apologies
in advance for the horrific typos and misspellings,
I think we're getting a higher and higher prepondeance
of fast talkers presenting these days.  :)

Matt



2008.02.20 Lightning talks

Three lightning talks, 10 minutes each.
(notes from all 3 in same file for me.)

Martin Hannigan
from start to transport: submarine cable deployment

hannigan at gmail.com, Keflavik Iceland
deployed multiple submarine cables, has a
background as a captain, done submarine cable
deployment.

iceland surrounded by water
lowcap cable <5G
hicap cable 720G
lowcap cable is to EOL soon
hicap subject to backhoe issue
Farice1

Drop a berrilium steel splitter in
middle of atlantic to peel off fibers
to islands.

Security benefits--communication to other
countries, rights to access, freedom of
information.
Performance concerns (speed of light, etc.)
.75ms per 150km
ROI--will you get return on your investment

parameters
wet/dry plant
manufacturer (tyco)
capacity
desktop study (DTS)
tends to be vendor locked in for full system.
biggest cable is 8pair cable.
undersea routing--where will you drop it

Risks
 length--too long may be too muc delay
 repeater distance and sparing; closer, the
  more you can pump out and handle upgrades to
  future tech, but costs more.
 shipping lanes, fisheries, geological considerations
  fisherman have associations; you've got to talk to
  the commercial fishers
 anchors, backhoes--more than 50% cause of outages.
 permitting to cross international boundries, and pay
  taxes.

sign intent to proceed; gets materials going
commision desktop study
realeastate for landing spots
complete engineering
ocean survey, map topography A to Z
manufacturer cable in segments, then join it at
repeaters.
A/Z ends dropped to the beach, cable rolled out

Suprise, Greenland connect!
3 routes out of country now

41 active underwater US east and west coast

50 active cable ships that lay/repair
ships are shared out, contracted to each other
if another ship is closer to where a cut is;
they all use a common jointing technique.
Cables powered from landing stations, 48vDC
powered into the cable.
More denser, the cheaper.

east coast to london, cables cross all over
the place.

Telegeography and Cable Operators are your friends.

Use armor on cables, bury offshore; costs more, but
worth it.

prefix mapping to sub cables might be interesting.

Be wary of FUD.

Fishing conflicts analysis--reroute around fisheries.

all nautical charts are in meters; 2150 meters down.

Bathymetry charts showing currents, can cause
friction

landing station approach, very shallow.

Credits.

wow.


========================================================


Duane Wessels, the measaurement factory
day in the life of the internet measurement project.
will have a measurement day in March.
our data will help researchers
You don't have to send data to them; just need it to
be collected, indexed, into database so researchers
can get to it.

motivated by 2001 report challenging reseachers to
collect a solid day's worth of data.

CAIDA and OARC did one in 2002, 48 hour DNS root
server collection

expanded in 2007

inspired to further expand for 2008

2007
5 root servers (CEFKM)
NaMeX internet exchange, AS 112
Open root servers, etc.

what did they learn?
5% of clients responsible for 95% of load
at root servers
1-2% of queries are legitimate; rest is junk.
query rates doubled from 2006
anycast very effective

shows where clients are vs where servers are
longitude chart vs bar colours showing where
clients connect to servers.

what is all the pollution at the roots?
lots of repeated queries, identical queries
invalid TLDs seem to hit F.root

what learned?
you always need more disk space than you think
collecting is easy; indexing is hard.

questions to pursue with DITL data: DNS
source of garbage
trends in v6, DNSsec, etc.

traffic/performance questions
botnets?
voice/video?
is v6 different than v4?
RND vs commercial traffic patterns?

need to accomodate privacy/legal concerns
Make data selectively available, anonymize it.
index it into CAIDA's DatCat

Please participate!

http://www.caida.org/projects/ditl/
http://imdc.datcat.org/


=====================================================


Peter Loesher
slide decks are all on the website, btw, so feel free
to read through thm on your own.

Joao_Damas at isc.org

BIND, AAAA and the root servers

finally, AAAA records for the root servers in the
root-servers.net zone
initially 6 out of 13 are providing v6 root service

about 80-100qps of v6 right now.

F root uses v6 in same way as v4; all local nodes
are advertising /48, covering /47 advertised on the
global nodes to avoid blackholding
13 of 43 nodes worldwide serving v6.
Most traffic going to paris and amsterdam, rest
goes to global nodes in bay area.  very little in
japan, possibly due to peering culture in japan.

peering with f root over v6
we are ready almost everywhere, let them know,
they'll peer with you via v6

BIND changes
BIND needs no code changes on its own.
ISC will provide updated hints file in 9.5
in the mean time you can fetch your own copy
from:
ftp://rs.internic.net/domain

*whew*

PGP signing party in Garden room, quick break.

Final talks will start at 11.



More information about the NANOG mailing list