Transcript Routing Measurements

Routing Measurements
Matt Zekauskas,
[email protected]
ITF Meeting 2006-Apr-24
Pointer to this presentation
• Right now:
http://people.internet2.edu/~matt/talks/20
06-04-24-SMM-ITF-RoutingMeasurements.pdf
• Soon also on the meeting pages
2
Matt Zekauskas
Overview
• Chris has provided motivation and some
examples
• I’m going to go through some tools that
can help, that we recommend installing if
you do not already have them
• I’ll provide some other pointers to
commercial tools I am aware of, but have
no experience with
3
Matt Zekauskas
RouteViews
• A project started at the University of Oregon
• In partnership with commercial carriers
• Original idea: low-cost Cisco router peers with
as many other networks as possible
• Each network provides it’s own view of the
routing table
• Thus, this central point has many views into
the routing table – Route Views.
4
Matt Zekauskas
RouteViews
• Today: multiple central servers, including the a
Cisco system and some PCs running Zebra
routing code
• Central implies BGP multi-hop, which means
the central server only gets the best view of a
given route from each source, and can’t get
information about a disruption if the disruption
also severs the link to the central collector
5
Matt Zekauskas
RouteViews
• Placing servers out at networks to catch
more routing anomalies and updates
during network outages
• Additional work to see more than just the
best route (Chris’ talk)
6
Matt Zekauskas
RouteViews
• Chris has shown the usefulness
• More views => A better picture => More insight
• The central collectors have additional capacity
to accept new peerings.
• Peer with a central collector – it’s easy and can
be done quickly.
• Consider adding a local collector if you have
the space and interest, but multi-hop BGP peer
as a first step
7
Matt Zekauskas
RouteViews
• http://www.routeviews.org/
• Contact to create a peering or prepare for a
local collector: [email protected]
•
•
•
•
IP Address
AS Number
Contact name & email
Any documentation on BGP communities
• Please let us know as well
([email protected], [email protected])
8
Matt Zekauskas
End-to-End Views
• RouteViews gives you snapshots of the
routing table -- how one network gets to
another.
• Sometimes you just want to know the
router hops from one point to another
• Verification of RouteViews information
• Quickly understand what’s happening now
• Constant monitoring of important points
9
Matt Zekauskas
End-to-End Views
• Two projects have been taking routing
measurements for some time, in the form of
traceroutes from measurement points
• AMP – general mesh among probes
• IEPM-BW – Mesh for high-energy physics
• In addition, the skitter project from CAIDA has
been doing much larger point-to-multipoint
measurements at low frequency for mapping
http://www.caida.org/tools/measurement/skitter
/
10
Matt Zekauskas
AMP
• Round-trip delay, loss
• Traceroute
• Many of you have AMP machines installed
• http://amp.nlanr.net/
• Currently the project is not emphasizing
expansion
• However, still willing to add machines (you
supply); there is also code available you could
install yourself for points of interest
11
Matt Zekauskas
AMP Example: new link
• In January, a TIEN2 link came up
between AARnet and the APAN Tokyo
XP.
12
Matt Zekauskas
13
Matt Zekauskas
14
Matt Zekauskas
• Week of 15-January-2006
15
Matt Zekauskas
16
Matt Zekauskas
17
Matt Zekauskas
AMP Example: large route change
• Amp-korea to amp-hutfi
(Taejon, KR/NOC to Helsinki, FI/Univ.)
• 16-April-2006 through 22-April-2006
18
Matt Zekauskas
19
Matt Zekauskas
20
Matt Zekauskas
21
Matt Zekauskas
AMP status
• Currently the main project is not focusing
on deployment.
• If you supply a machine, they are willing
to add it to one of the main meshes
(general, international-only).
• Contact Tony McGregor,
[email protected]
22
Matt Zekauskas
AMP Software
• You can also install your own mesh:
http://amp.nlanr.net/Software/AMP/
• The central collection software has not
yet been released, but will be “soon” (will
give out on request now)
• Tony also has some “gross change”
detection code that he said may be
available end-year.
23
Matt Zekauskas
PingER (& IEPM-BW)
• Point-to-multipoint round-trip delay, loss
• But no traceroutes, for that IEPM-BW
• From beacons at major high-energy physics
sites, to many other high-energy physics sites
• Including many that were originally not on
“research networks”
• http://www-iepm.slac.stanford.edu/pinger/
24
Matt Zekauskas
IEPM-BW
• A mesh of throughput tests for high-energy
physics (think upcoming LHC experiments)
• One-to-many focus
CERN, Brookhaven, Fermilab, SLAC, Caltech,
and a site in Pakistan.
• In addition, takes traceroutes among the probe
sites
• http://www.slac.stanford.edu/comp/net/iepmbw.slac.stanford.edu/slac_wan_bw_tests.html
25
Matt Zekauskas
IEPM-BW Example
26
Matt Zekauskas
On-Demand Tools
• Two tools are useful to understand
routing right now
• Traceroute servers
• Routing from a given point, as detected by
traceroute
• Looking glass servers
• Show a particular routing table entry right now
• Actual, versus inferred, network connectivity
• But only shows next hop
27
Matt Zekauskas
Traceroute Servers
• Servers that provide on-demand
traceroutes from specific points.
• Many networks have at least one
• Abilene allows traceroute from our
routers via the IU ‘router proxy’
http://ratt.uits.iu.edu/routerproxy/abilene/
28
Matt Zekauskas
Traceroute Servers
• List of HEP servers, pointer to decent server
source code, and pointers to lots of other lists
of servers:
http://www.slac.stanford.edu/comp/net/wanmon/traceroute-srv.html
• Lots of other servers: www.traceroute.org
• Presented graphically:
http://www.caida.org/analysis/routing/reversetr
ace/
• The last two have both traceroute and looking
glass lists.
29
Matt Zekauskas
Traceroute Servers
• Very useful – install one of these and
publish it’s location on your NOC page
• Unless, you install a looking glass server
that also has traceroute functionality…
30
Matt Zekauskas
Looking Glass Servers
• Similar, for routes at a particular point
• http://www.traceroute.org has a list
and pointers to other lists & source code
• Often (always?) contains traceroute server
functionality
• Again, the Abilene router proxy can be used in
the same way
• Recommend installing one of these (as well),
and publish its location
31
Matt Zekauskas
Some Commercial Tools
• Mainly to help you understand your own
network, don’t give a global view
• Packet Design’s RouteExplorer
http://www.packetdesign.com/products/pr
oducts.htm
• IPsum Networks’ Route Dynamics
http://www.ipsumnetworks.com/
32
Matt Zekauskas
Communicate
• If you discover something odd, you would like
to be able to understand and resolve the
problem (if one exists)
• Place tools to help others
• Utility grows with the number of installations
• Publish tool links on your NOC pages
• If you see a problem, contact appropriate NOC
• Jim Williams, [email protected], is interested in
facilitating NOC-to-NOC communication
33
Matt Zekauskas
34
Matt Zekauskas