Transcript OWEN/NERO Bandwidth Audit

1
Internet2 and Oregon
Southeast Technology Consortium I2 Day
May 9, 2001
Joe St Sauver, Ph.D.
([email protected])
Computing Center
University of Oregon
2
Background and Introduction
• I’m the Assistant Director for Academic
User Services, here at the University of
Oregon Computing Center in Eugene, and
I’d like to thank Bonnie Neas, Assistant VP
for Federal Government Relations and
Director of Internet Research at NDSU for
inviting me to talk to you today. Hopefully
the H.323 technology will work well, and
hopefully I won’t be called for jury duty. :-)
3
What My Group and I Do
• I run a sort of unusual group here at the CC.
In addition to supporting academic users of
our large systems and microcomputers, my
group and I also produce documentation,
maintain the UO home page, run network
service boxes (such as Usenet News servers,
web cache boxes, ftp archive servers, video
servers, etc), handle network abuse and
policy issues... and we work on advanced
networking issues, such as Internet2.
4
What We’re Going
to Cover Today
• The history of Oregon and Internet2…
• Some technical areas we’re interested in
and working on...
• And some opportunities for collaboration
between Oregon and North Dakota.
5
A Brief History Of Internet2
in the State of Oregon
Or…
“How in the world did UO end up
collaborating with/doing/sponsoring
all these Internet2-related things?”
6
Long Long Ago (Back ~1995),
Before There Was Internet2...
• Oregon had a nice statewide network, the
Network for Education and Research in
Oregon (NERO), with hub sites in Eugene,
Corvallis and Portland, interconnected by
OC3 (155 Mbps) and DS3 (45 Mbps)
circuits. A nice overview of the early days
of NERO, written by Dave Meyer, the guy
who built it, is available online at
http://sith.maoz.com/~dmm/i2days/
7
NERO’s Role
• NERO served as the production backbone
for Oregon higher education, and also
provided a platform for research on the
network, and research via the network.
• It was one of the earliest statewide ATM
networks (it’s now packet over sonet)
• It provided both intrastate connectivity and
commodity Internet transit service via
UUNet and MCI (now CWIX).
8
Oregon Internet Exchange
• Meyer also built out the Oregon Internet
Exchange (OIX) at UO, a network meet
point where ISPs could come to exchange
customer network traffic without paying
financial settlements, thereby keeping local
traffic local and reducing bandwidth costs.
See http://www.oregon-ix.net/
• We’ll talk about how the OIX fits into the
larger I2-in-Oregon picture later in this talk
9
The Original vBNS
• In April of 1995, the vBNS became
operational. It was originally deployed as a
way of providing connectivity between
federally funded supercomputing centers at
Cornell, Pittsburgh, San Diego, NCAR and
NCSA/UIUC, and the four NSF funded
network access points. See, for example:
http://www.vbns.net/presentations/
krnet-tutorial/index.html
10
The Revised vBNS and the
New Connections Program
• A presidential review committee looked at
the original vBNS, and basically said,
“Great, world-class, network. Too bad no
one’s using it.”
• Ergo, the impetus for the NSF’s New
Connections program, and expanded access.
11
The Original UO/OSU NSF
New Connections Application
• As part of his work with NERO, Dave
Meyer, in conjunction with researchers from
Oregon State University, applied for an NSF
Connections Grant program grant in 1996.
Approval would mean funding, plus
eligibility to use the vBNS (very cool!)
• That proposal envisioned that UO and OSU
would share a single vBNS DS3 (45Mbps),
interconnecting the site via NERO.
12
We Won...
• Our application was approved by the
NSF in December of 1996 (see:
http://www.fastlane.nsf.gov/servlet/
showaward?award=9617043).
13
… And We Lost
• However the NSF decided, post hoc, that in
fact a separate grant was needed for each
school to be connected to the vBNS, even if
we planned to share a common connection.
• UO and OSU jointly agreed that Oregon
State would take the original grant money
and the initial vBNS DS3 connection, and
UO would reapply during the next round.
14
And Then We Lost
Some More...
• We had expected the reapplication to be a
relatively routine process and to only cause
a short delay, but then funding for the NSF
Connections Grants program got tangled up
with the Intellectual Infrastructure Fund (the
so-called “DNS registration tax”), and the
NSF Connections Program was temporarily
put on hold.
15
But Then We Finally
Won (Again)
• Finally, in September of 1998, UO was
approved to connect to the vBNS (see:
http://www.fastlane.nsf.gov/servlet/
showaward?award=9729628
16
Having FINALLY Won...
• ... we found the world a different place.
• The vBNS had basically priced itself out
of the market, and Abilene was then the
clear network of choice for new connectors.
• ATM had died the death it richly deserved.
• Gigapops (like hula hoops) were the big
thing (everyone had to have one).
17
The Changing World (cont.)
• OC3 connectivity was suddenly affordable.
In fact, changes in prices made it possible
for us to get not one, but two Abilene OC3
connections. That multihoming gave us
connection diversity and survivability.
[We believe that Oregon was the first
multihomed site on Abilene.]
18
But Even After We Got
Approved, We Weren’t Done
• Our “getting connected to Internet2” saga
actually has many more twists and turns.
• For example, we almost lost our
backhauled connections via the Qwest
backbone when USWest was slow to deliver
our OC3 local loops, and the Oregon
Legislature wanted a NERO bandwidth
audit (see https://web-vms.uoregon.edu/
~joe/bw2/owen/index.html)
19
Bottom Line: Oregon Did
Finally Get Connected
• But when all was said and done, we DID
get connected. We ended up with the
Oregon Gigapop (OGIG) providing
connectivity for the University of Oregon
via two Abilene OC3s connections, one to
Denver and one to Sacramento (now
Sunnyvale).
• It look us a while to get there, but we DID
get there.
20
And By Having To Wait, We
Ended Up Really Winning
• We ended up with 310 Mbps instead of 45
• We ended up with the Oregon Gigapop
located in Eugene
• And by the time we got on Internet2, I2 was
far more practically useful than it had been
earlier (remember that the value of a
network is proportional to the square of the
number of entities it connects)
21
And We Made Some Good
Friends While We Waited...
• For example, following a mandate from the
state government that Oregon public
telecommunications networks should be
aggregated to the maximum extent possible,
OPEN (the Oregon Public Education
Network, the network serving public K12
institutions in Oregon) joined NERO, and
the Oregon state government agency
network also joined NERO.
22
NTIA and K12
• It was also during this period that UO began
working with K12 institutions in the south
half of the state as part of an NTIA grant
program (41-40-94029), creating Lane
Education Net (LEN), which eventually got
merged into OPEN when the networks
consolidated.
23
And With That Background,
The Stage Was Set...
• Institutions in the state had I2 connectivity
• Grant funding for those connections was
only for a limited time… tic, toc, tic, toc...
• We had a solid intrastate network
connecting the state universities, K12 and
state government (including providing
commodity Internet for those partners)
24
The Stage Was Set (continued)
• We’d already been working with K12, both
via OPEN and as part of the NTIA grant
• We had a commodity Internet exchange
point
• There was a tremendous Internet2 public
relations effort, generating a lot of
institutional, governmental and public
interest
25
Evolution 1:
OSU Leaves The vBNS
• Along the way, Oregon State had gotten a
supplemental grant to provide additional
funding for their vBNS connection, but
when that eventually ran out, we began
tunneling their Internet2 traffic via NERO
to the Oregon Gigapop.
• Now OSU pays for one of the two Oregon
Gigapop OC3s outright.
• OSU represented OGIG’s 2nd participant.
26
Evolution 2: PREN in Portland
• In August of 1999, a consortia of Portland
area schools (PSU, OHSU and OGI),
known collectively as PREN, applied for an
NSF Connections Grant, proposing to
connect PREN via donated WCI OC12
connectivity backhauled to UW’s gigapop
in Seattle. PREN’s application to the NSF
was approved (http://www.fastlane.nsf.gov/
servlet/showaward?award=9975992)
27
And About The Portland
Schools Going North...
• We won’t BS you: there was significant
concern/angst/debate about the Portland
school’s decision to connect via Seattle,
rather than via the Oregon Gigapop.
• At the same time, we completely
understood that it would be very hard to
turn down a “free” OC12.
• Besides, it wasn’t an either/or situation, at
least for Portland State.
28
Portland State Was In a Unique
Position
• Like UO and OSU, PSU was a long time
NERO partner and already had OC3 class
intrastate connectivity.
• They could have connectivity via PREN
and the UW gigapop in Seattle, but they
could also have connectivity from the
Oregon Gigapop via NERO.
29
Portland State:
Turned Up Immediately...
• So, as soon as the NSF approved PSU for
connection to Internet2, the Oregon
Gigapop was immediately positioned to
provide I2 connectivity for them, as we
continue to do to this day.
• PSU thus became OGIG’s third participant.
30
Evolution 3: The I2 Sponsored
Participant Program
• Meanwhile, Internet2 had announced an
interest in accommodating institutions
(called “sponsored participants”), which
didn’t fit the traditional Internet2 profile
(for instance, K12’s, non-Carnegie Research
1/Research 2 schools, etc.).
• We were immediately intrigued.
31
We Submitted An Application
for “3” Sponsored Participants
• UO immediately submitted an application
for “three” sponsored participants:
-- the Oregon University System colleges
which weren’t yet Internet2 members,
-- Oregon’s public K12 network (OPEN),
(representing about 600K users), and
-- Oregon’s state government agency
network.
32
Onesie-Twosies vs. Groups
• Apparently our application for the “three”
sponsored participants were something of a
bombshell within Internet2, even though
we’d alerted I2 to our anticipated approach,
and at least some Internet2 staff had said
that they thought that that would be fine.
• Apparently Internet2 had been planning to
handle sponsored participants on a onesietwosie basis, rather than on a group basis.
33
Why Did We Apply To Sponsor
Groups of Sites?
• We already worked with OPEN and the
state government and the other state
colleges as group entities. We didn’t want to
disaggregate them just for the purpose of
applying for I2 connectivity for them.
• We felt it was very important to respect
existing organizational structures and
existing network support structures.
34
Why Do Groups? (continued)
• We also felt that it would be a bad thing
from an equity point of view to promote
Internet2 to some Oregon K12 schools but
not to others, or to promote it to some
NERO partners, but not to others.
• There was no question that technically
networks should interconnect at the
Autonomous System level. Each of the
three groups was an ASN-level entity.
35
Why Do Groups? (continued)
• There was also the issue that many K12
schools, at the school level, do not have the
funds to provide on-site wide area network
support staff, nor do they have the money to
buy separate dedicated connections.
• Doing only some schools but not others
would be confusing, and limit opportunities
for collaboration, tech transfer, and selfsupport among Oregon schools
36
Why Do Groups? (continued)
• The paperwork burden would also be
essentially the same for any of these:
-- all K12 schools in the state of Oregon, or
-- a single educational service district, or
-- a single school district, or
-- a single school, or
-- a single school building, SO...
• You clearly get the most “bang for your
buck” by adding an entire K12 statewide
network at a time
37
The Oregon University System
Regional Universities
• While the sponsored participant program
was getting sorted out within I2, Internet2
did agree to let us reapply to sponsor four
individual colleges -- Eastern Oregon
University, Oregon Institute of Technology,
Southern Oregon University and Western
Oregon University -- on a onesie-twosie
basis. They were approved and hooked up
to the Oregon Gigapop via NERO.
38
What About K12? Tah Dah...
Sponsored Groups (SEGP)
• After much process and discussion,
Internet2 decided to allow connection of
state K12 networks as Sponsored
Educational Group Participants (SEGP’s).
• Oregon’s statewide K12 network, OPEN,
was one of the first five approved (along
with Michigan, Missouri, Virginia and
Washington State). We are currently
completing OPEN’s SEGP paperwork.
39
Interesting SEGP’isms...
• New fee ($30K + $2K * # of US Reps)
• SEGP’s can’t include Carnegie Doctoral
Research Extensive (old “R1/R2”) schools,
nor former members of Internet2
• Internet2 can’t be used to provide intrastate
connectivity for SEGP institutions (e.g., no
competition with state backbone networks)
• Each SEGP must have its own ASN
• No new PVCs or BGP peerings for Abilene
40
Who In Oregon Is Still
Not At the Oregon Gigapop?
• The smaller private liberal arts colleges in
the state aren’t connected to I2 or to NERO
• Oregon community colleges aren’t part of
NERO, and hence we don’t have a clean
way of bringing them in as a group.
• State government, while part of NERO, still
isn’t eligible for connection to Internet2.
• County and municipal governments….
41
About the Smaller Private
Liberal Arts Colleges...
• Currently the focus of the latest NSF
connections funding program (NSF 01-73)
• The tricky part is doing the local loops or
intrastate network to get them connected,
and the money available ($150K/biennium)
may not be enough, particularly when you
figure in the likely need for capital
investments in hardware for each school
42
Oregon’s Community Colleges
• Tough to partner with them as a group
because they have traditionally purchased
commodity Internet connectivity directly
from commercial network service
providers; there is no existing intrastate
community college network per se
• Some are in geographically challenging
locations (e.g., Treasure Valley CC in
Ontario, Oregon, near the Idaho border)
43
What About State
Government?
• State government was denied participation
outright (truly unfortunate, given resources
like the state library, state museums, and
various research labs), but c’est la vie.
• State government COULD be a real ally
when it comes to leveraging assets and
providing funding
• An odd disconnect has the potential to occur
44
County and Local Government
• County and local government agencies
aren’t on anyone’s radar right now, but they
should be. Why?
• Who owns all that right of way that fiber
companies need? Yep, the cities and
counties do. Who owns all that right of way
that maybe YOU’D like to use for your own
fiber? Yep, the cities and counties do.
• You ignore/forget about your friends in
county and local gov’t at your peril.
45
County and Local Government
and Oregon Fiber...
• http://www.teleport.com/~samc/
telecom/index6.html
• http://www.ruralfiber.net/
• http://www.bpa.gov/Corporate/KCC/
conferences/fiber/agenda.htm
• And for a brief tutorial introduction to fiber:
http://cc.uoregon.edu/cnews/summer2000/
cnsum2000.pdf at page 18.
46
Why Did We Bother Telling
You This I2 Oral History?
• It explains, in part, why we work so hard to
share what we’ve been able to build.
• It may give you hope: while it hasn’t been
easy getting folks connected in Oregon, we
HAVE made progress, and you can, too.
• It may help you to understand some of what
we do, have done, and will do. (It also clearly
shows that we’re not done yet.)
47
Some Technical Areas We’re
Currently Working On
“How did you pick areas to work on?”
“We tried everything, and dropped the
stuff that seemed not to work very well.”
48
Some Examples of Areas We’re
Currently Working On
•
•
•
•
•
•
•
•
IP Multicast (including SSM)
IPv6
End to End Performance
Network Measurement
Usenet News
Less-Than-Best Effort Service
I2 Traffic and Commodity Internet Traffic
Routing Issues
49
IP Multicast
• While most Internet traffic is “normal” IPv4
unicast traffic, with bits getting sent from
one sender to one receiver, there is an
additional type of traffic called “IP
multicast.”
• IP multicast allows traffic to be efficiently
shared among multiple receivers, if your
network is configured to support it (as
Abilene does, and as your campus can...).
50
Why Is IP Multicast
Important?
• IP multicast is important, for example,
because it allows institutions to deliver high
quality (640x480x30 frames per second)
MPEG1 video at 1 to 1.5Mbps to large
audiences without crushing the network -one viewer or ten thousand, the load is the
same.
• “It just looks like TV on my computer!”
51
The Downsides to IP Multicast
• High bit rate content (1-1.5Mbps) is still too
fast for off-campus users; dialup, cable
modem and DSL users need not apply (but
1-1.5 Mbps is not an issue on-campus)
• IP multicast deployment has developed an
undeserved reputation for being “hard”
• Because there aren’t many viewers, there
isn’t much content; because there isn’t
much content, there aren’t many viewers.
(But we’re working on fixing that.)
52
What Can I See Via
IP Multicast Today?
• We currently offer a variety of MPEG1
content across Abilene and the commodity
Internet today.
See, for example:
http://cc.uoregon.edu/cnews/spring2001/
spring2001.pdf (pps. 4 and 13) and see
http://videolab.uoregon.edu/
53
SSM (Single Source Multicast)
• SSM is important because it makes it
possible to cleanly integrate multicast
content into web pages, removing issues
associated with sdr and sap.
• SSM requires IGMP v3, which means that
deployment is going to be a gradual (rather
than overnight) phenomena, but it only
requires IGMP3 on the last hop router. See:
ftp://ftpeng.cisco.com/ipmulticast/ssm/
index.html
54
For More Information
About IP Multicast
• An excellent starting point for training
technical folks about IP multicast is the
Cisco IP multicast training materials
available online at:
ftp://ftpeng.cisco.com/ipmulticast/
training/index.html
55
IPv6
• Another important area that is beginning to
get more press (now that IP multicast has
become more-or-less a routine item on
Abilene) is IPv6.
• IPv6 is designed to preclude the possibility
of address space exhaustion by replacing
current 32 bit IPv4 addresses with 128 bit
addresses. It is already quite popular in
parts of Asia, where addresses are scarce.
56
How Does IPv6 Fit Within I2?
• IPv6, like IP multicast, is an excellent
example of an advanced network service,
literally a next generation network,
which is supported over Internet2.
• And, just like IP multicast, while IPv6 can
be done without Internet2, having Internet2
available makes it far easier to begin
experimenting with IPv6.
57
IPv6 Action Items
• If you’re interested in pursuing IPv6, the
laundry list to begin working on includes:
-- arrange for tunneled IPv6 connectivity
-- arrange for temporary IPv6 address space
-- get local DNS service in shape to handle
AAAA IPv6 address records
-- get host operating systems IPv6-ified
-- get applications IPv6-ified
-- spread the word to your friends. :-)
58
For More Information
About IPv6
• See http://cc.uoregon.edu/cnews/
spring2001/spring2001.pdf at pages
13 to 16.
59
End-to-End Performance
• Currently, while all I2 sites have large
pipes, individual users on typical systems
have a hard time obtaining more than three
to six Mbps transfer rates due to a variety of
factors, which frustrates users. Most users
have an expectation that if (a) their
institution has a large pipe to I2, and (b)
they have fast ethernet to their desktop, then
they should naturally be able to go “fast.”
60
But In Fact Utilization of I2
Connections Can Often Be Low
• This has motivated an Internet2 End-to-End
performance initiative, aimed at improving
single stream end-to-end throughput
• Put plainly, I2 needs to figure a way to help
users at Internet2 schools use their large
connections more effectively. If pipes stay
largely idle, Internet2 will have a real
problem.
61
For More Information
on E2E...
• http://www.internet2.edu/e2e/
• http://www.internet2.edu/e2e/
planningmeeting.shtml (including a
onepager I wrote)
• http://darkwing.uoregon.edu/~joe/
how-to-go-fast.ppt
62
Measurement
• Another important area is that of network
performance measurement
• Oregon participates in a number of
measurement projects, including NLANR’s
AMP project, Advanced.Org’s Surveyor
project, Vern Paxson’s NIMI project,
CAIDA’s Skitter project (we’re the only
higher ed institution in North America to
host a Skitter box), etc.
63
Why Is Measurement
Important?
• Achieving sustained high throughput
requires a virtually lossless network, and
you need to be able to isolate and identify
any network problems which arise
• Because Internet2 crosses organizational
domains and uses a variety of network
technologies, changes (for better or worse)
happen all the time.
64
Nice Overview of
Measurement Initiatives
• A nice overview of Internet measurement
tools is available from CAIDA at
http://www.caida.org/tools/taxonomy/
• Every site should consider participating in
one or more measurement initiatives, if at
all possible. [In most cases, it is relatively
easy to host a measurement box, with the
worst hassles often associated with getting a
GPS antenna onto a roof for time sync’ing.]
65
Route Views
• The University of Oregon also provide a
route viewer, with views of the Internet
BGP routing table from fifty different
network service provider/location
combinations. See:
http://www.antc.uoregon.edu/
route-views/ and
http://rv-archive.uoregon.edu/
66
Usenet News
• Another area where UO has been very
active on Internet2 has been in the area of
Usenet News. Usenet News is both a
traditional production network service of
some importance, and also a service that is
extremely well suited to Internet2’s unique
architecture.
• We currently do Usenet News feeds with
virtually all I2 schools that do Usenet News.
67
Usenet News Traffic Volume
• Usenet News has also been identified as a
major application in terms of Internet2
traffic volumes; see:
http://darkwing.uoregon.edu/~joe/
how-to-go-fast.ppt at pages 14-15, 38-40.
• Shifting that traffic from commodity
Internet connections to I2 connections
represents a real savings for institutions.
68
Weather Data Via Usenet News
• We also have been working with
NCAR/UCAR to see if Usenet News is
adaptable to transportation of LDM weather
data files; see, for example:
http://www.unidata.ucar.edu/staff/anne/
nntpReport.html
69
IPv6, Multicast and Usenet
• Usenet News also provides a nice vehicle
for testing IP multicast data delivery (see:
http://www.newscaster.org/) and a nice
“foundation tenant” application for IPv6
connections.
70
Best Effort Service vs. Other...
• The current model of undifferentiated
delivery of all types of network traffic
(email, web pages, IP video, voice over IP,
etc.) is called “best effort service.”
• There has been substantial interest in
offering a “better than best-efforts service,”
a so-called premium delivery service (ala
Fedex for package service) via Quality of
Service (QoS) mechanism.
71
Problems That the QoS
Premium Service Has Run Into
• So far, it has proven to be impractical to
implement premium service via QoS for
inter-domain network traffic, because:
- there’s no way to police the edges of the
net to block unauthorized premium traffic
- the network is currently uncongested, so
there’s no need for differentiated service
- doing bandwidth reservations, doing user
authentication, and doing billing is a pain
72
Less-Than-Best-Efforts
“Scavenger” Service
• Along the way, some Internet2 folks have
considered offering a less-than-best effort
service, a “scavenger service,” that would
be the antithesis of premium service,
targeted at delay-insensitive or low-priority
traffic. See: http://qbone.internet2.edu/qbss
• It is still an open question whether that
effort will result in a usable/used service
73
Issues for Scavenger Service
• The network core is uncongested, so the
motivation for deprioritizing traffic is…?
• Scavenger service is only available for I2,
but where it may be really needed is on the
commodity Internet
• Unclear how traffic would be selected for
marking with the SS code point, and how
the marking would actually occur.
74
Speaking of Traffic
Management
• We have also been interested in the
relationship between Internet2 and
commodity Internet traffic for some time.
• We have postulated that most users won’t be
able to routinely ascertain when they are
using I2 and when they aren’t, so if you
encourage users to go fast anywhere, you
need to support them going fast everywhere.
75
Why Is Going Fast Everywhere
Potentially a Big Deal?
• While high bandwidth connections to
Internet2 are relatively inexpensive (no,
really, they are, at least by comparison!),
commodity Internet connections continue to
be an order of magnitude more expensive
(e.g., you might pay a $110K per YEAR for
an Abilene OC3, but $100K per MONTH
for a commodity OC3 from a major network
service provider).
76
And That’s Why Exchange
Points Become Important...
• Remember we mentioned the Oregon
Internet Exchange earlier -- this is where
you begin to see the crucial role it will play
for Oregon -- the Oregon IX will allow us to
minimize the amount of commodity Internet
transit bandwidth we need to purchase,
substituting inexpensive peering for at least
some network destinations.
77
And Exchange Point Issues
Motivate Fiber Initiatives
• Unfortunately, running an exchange point in
Eugene is harder than running an exchange
point in Portland (which is still harder than
running an exchange point in Palo Alto,
say,) simply because there are fewer
providers available to peer with in Eugene...
• QED, eventually you need to get
“inexpensive” connectivity to “larger”
cities to continue to add network peers.
78
Route Asymetry
• We’ve also been interested in the issue of
route asymetry. Routing asymetry arises
when traffic goes out via Internet2, but
comes back via a non-Internet2 network
connection (or vice versa). In some cases
the asymetry is intentional and due to
differences in connection costs or
connection capacity; in other cases it is just
a matter of routing being broken.
79
A Hypothetical Example
of Routing Asymetry
• Assume we have three possible paths from
us to Oregon Graduate Institute in Portland:
(i) via I2 (Eugene down to Sunnyvale, back
up to Seattle, down to Portland); (ii) via
peerage with Verio at the Oregon IX
(virtually direct, and traffic carried at no
cost to us); or (iii) via the commodity
Internet. Which is the “right” or “best” path
for that traffic to take?
80
Thinking About the
Available Paths
• If we pref traffic to OGI to go via the
Oregon IX, it would go to OGI via a direct
route, at no cost to us and with low latency
(and low latency ==> faster throughput).
• But… what if OGI is a customer (e.g., buys
commodity transit) from Verio, and hence
would prefer its traffic from us to go to via
its lightly loaded I2 connection, even if that
traffic follows a less-direct path?
81
Why Not Just Always Take the
I2 Path, If It Is Available?
• Taking the I2 path by default is sometimes
inefficient at best. For example, consider
UO to Portland State, both I2 member
schools, AND both connected to NERO.
Should we REALLY be preferring the I2
path between those schools to the direct
path over our intrastate network, NERO?
No, of course not, that makes no sense.
82
Nice Routing Asymetry
Study By Hank Nussbacher
• See: “The Asymmetry of Internet-2”
http://www.internet-2.org.il/i2-asymmetry/
index.html
83
Opportunities For
Collaboration
“Collaboration is, in general, a whole lot
easier today than it was during WW II.
For example, now you don’t get shot just
because you’re a collaborator.”
84
A Fundamental Truth
• Networks are only useful when they
connect somewhere you need to go. In most
cases, the places you need to go are where
your friends are…
• “But what if I don’t know anyone anywhere
else in Internet2 to work with?”
85
Joe’s Matchmaking Service
• For better or worse, one of the things that
has resulted from our work with Internet2 is
that we know folks all over the place.
• I’ll try, to the extent I’m able, to help get
you connected with folks at other sites who
might be interested in working on network
projects of common interest. Send me email
with information about a project you’re
interested in pursuing, and I’ll try to help.
86
Internet2 Working Groups
• Another good way to meet colleagues in
Internet2 is via I2 working groups. Most of
the working groups do business via email
mailing lists, plus periodic working group
meetings at I2 member meetings or
NLANR/I2 Joint Tech meetings. For more
information, see:
http://www.internet2.edu/html/
working-groups.html
87
And If All Else Fails...
• We’re always glad to work directly with
folks on projects in cases where no one else
may be interested.
• We’re crazy guys, but we generally have a
lot of fun.
88
“But What If We Don’t Have
A Project In Mind?”
• Maybe consider one of the areas outlined in
section two of this talk?
• How about doing something with H.323
video, as used for this presentation?
• Maybe looking at peer-to-peer applications
would be of interest?
• Voice over IP?
• Security?
A Couple of Concluding
Thoughts...
• Internet2 is a tremendous resource, and you
CAN work successfully with it. You may
need to be persistent, but you can prevail.
• There are lots of interesting projects going
on in Internet2, including many areas where
you can get involved.
• We’d be glad to work with you, or to help
you to make connections if we can do so.
• Thanks for the chance to talk to you today!
89