Transcript Nakao-open-AFI2008summer

Research Activities
PlanetLab/GENI
Aki Nakao
Univ. of Tokyo / NICT
AsianFI 2008 Summer School
1
Outline
PlanetLab
GENI
Similar Research Activity in Japan and Asia
AsianFI 2008 Summer School
2
PlanetLab
The largest and most popular overlay network testbed
Currently consists of 863 nodes at 447 sites
800+ Projects/ 1000+ Researchers
JGN2/NICT collocate overlay nodes in Japan
Several Universities in Japan have joined
863 nodes
447 sites
40+ countries
1000+ researchers
800+ Projects
AsianFI 2008 Summer School
3
Brief History of PlanetLab
PlanetLab 1.0 (2002-3)
UNIX account slivers
PlanetLab 2.0 (2003-4)
Vserver slivers
PlanetLab 3.0 (2004-2006)
PLCAPI 1.0
PlanetLab 4.0 (2007-)
MyPLC1.0
Federation Idea
PlanetLab 4.2 (2008)
We are here
Federation (PLE, PLJ, (PLK, PLC,=>PLA?))
New Development (RSpec/GENI-wrapper, New Node Type,
Monitoring, QA,VINI)
AsianFI 2008 Summer School
4
New Generation Network and Overlay Network Test-bed
What is Overlay
Network Testbed for ?
Innovation
Creation of Ideas Feedback Cycle
Limitations in Prior Models
Evaluation at Labs
Publish
Evaluation at Planetary Scale
Acquire Users (Pilot Services)
Overlay Test-beds
Enables These
Our Goals
Establish Innovative Business
Contributions to Society
New Generation Network
AsianFI 2008 Summer School
5
CoDeeN: Partial Replication CDN
CCC.COM
B
→ PlanetWorks (2007-)
→ Coblitz(2007-)
A
BBB.COM
A
A
C
C
B
C
B
B
A
C
AAA.COM
client
server surrogate
redirector
Large File Transfer! Faster than BitTorrent!
AsianFI 2008 Summer School
6
What made PlanetLab successful?
Centralized trust
Centralized resource control
Decentralized management (unbundled management)
The bandwidth as free
Provide only best-effort service
Make Linux the execution environment
Do not provide distributed OS services (up to users)
Evolve the API
[Tom Anderson, Timothy Roscoe, “Learning from PlanetLab”, WORLDS’06]
Self-Scaling Architecture
[Aki Nakao]
AsianFI 2008 Summer School
7
Self-Scaling Architecture (as in P2P)
Consortium
Loosely bound
Membership Agreement
Donate a few resources
• At least two machines (minimum H/W requirement)
• Bandwidth
“Tit-for-Tat”
• Resource reciprocity
The more join, the more useful the system becomes
Primary incentives in P2P applications
AsianFI 2008 Summer School
8
PlanetLab Architecture
Stanford
PlanetLab
Node
Cambridge
PLC
(PlanetLab Central)
Resources
U.Tokyo The Internet
Princeton
resource
management
U.C.Berkeley
CMU
MIT
PlanetLab Architecture
AsianFI 2008 Summer School
9
PlanetLab New Generation
PlanetLabNG = GENI Prototype
PlanetLab 4.2 + geniwrapper
PLC wrapper: prototype done, integration underway
NM wrapper: prototype in progress
Wrapper includes…
interfaces
namespaces
security mechanisms
Migration plan
seed registries from PLC’s DB
Current and new interfaces coexist
unbundle PLC over time
[PlanetLab Developer's Meeting
May 13-14, 2008 (Princeton University)
experiment with peering
Overview, Deconstructing PLC and Meeting
10
Notes (Larry Peterson)]
AsianFI 2008 Summer School
GENI (briefly) with courtesy to GPO
Global Environment for Network Innovation
How we use GENI, how we build GENI?
GENI Spiral 1 / Prototype
Disclaimer: I have been granted permission to use
these slides from Chip Elliott and Aaron Falk
And present these on behalf of them so that we share
what is going on outside Asia.
AsianFI 2008 Summer School
11
The GENI Vision
A national-scale suite of facilities to explore radical designs
for a future global networking infrastructure
Virtualized
Deeply programmable
Programmable & federated, with end-to-end virtualized “slices”
Sensor Network
Federated
International
Infrastructure
AsianFI
2008
Mobile Wireless
Network
Summer School
Edge
Site
Heterogeneous,
and evolving over time via
spiral development
12
How to use GENI: A bright idea
I have a great idea! The original Internet
architecture was designed to connect one
computer to another – but a better
architecture would be fundamentally based
on PEOPLE and CONTENT!
That will never work! It won’t
scale! What about security? It’s
impossible to implement or
operate! Show me!
AsianFI 2008 Summer School
13
Trying it out
My new architecture worked great in
the lab, so now I’m going to try a
larger experiment for a few months.
And so he poured his
experimental software into
clusters of CPUs and disks,
bulk data transfer devices
(‘routers’), and wireless access
devices throughout the GENI
He uses a modest slice of GENI, sharing its facilities with
many other concurrent experiments.
suite, and started taking
AsianFI 2008 Summer School
14
measurements . . .
It turns into a really good idea
Boy did I learn a lot! I’ve published papers,
the architecture has evolved in major ways,
and I’m even attracting real users!
Location-based social
networks are really cool!
His experiment grew larger and
continued to evolve as more and
more real users opted in . . .
His slice of GENI keeps growing, but GENI is still running
many other concurrent experiments.
AsianFI 2008 Summer School
15
Experiment turns into reality
My experiment was a real success, and my
architecture turned out to be mostly
compatible with today’s Internet after all –
so I’m taking it off GENI and spinning it
out as a real company.
I always said it was a good idea, but way
too conservative.
AsianFI 2008 Summer School
16
Moral of this story
GENI is meant to enable . . .
Trials of new architectures, which may or may not
be compatible with today’s Internet
Long-running, realistic experiments with enough
instrumentation to provide real insights and data
‘Opt in’ for real users into long-running experiments
Large-scale growth for successful experiments, so good ideas
can be shaken down at scale
A reminder . . .
GENI itself is not an experiment !
GENI is a suite of facilities on which experiments run
GENI creates a huge opportunity for ambitious research!
AsianFI 2008 Summer School
17
GENI Spiral 1 has now begun!
First results expected in 6-12 months
GENI Project Office Announces $12M for
Community-Based GENI Prototype Development
July 22, 2008
The GENI Project Office, operated by BBN Technologies, an advanced
technologies solutions firm, announced today that it has been awarded a three
year grant worth approximately $4M a year from the US National Science
Foundation to perform GENI design and risk-reduction prototyping.
The funds will be used to contract with 29 university-industrial teams
selected through an open, peer-reviewed process. The first year funding will be
used to construct GENI Spiral 1, a set of early, functional prototypes of
key elements of the GENI system.
AsianFI 2008 Summer School
18
1st GENI Solicitation – proposal areas
AsianFI 2008 Summer School
19
Generous Donations to GENI Prototyping
Internet2 and National Lambda Rail
Internet2
10 Gbps dedicated bandwidth
National Lambda Rail
Up to 30 Gbps nondedicated bandwidth
40 Gbps capacity for GENI prototyping on two national footprints
to provide Layer 2 Ethernet VLANs as slices (IP or non-IP)
AsianFI 2008 Summer School
20
GENI Engineering Conferences
Meet every 4 months to review progress together
3rd meeting Oct. 28-31, 2008 in Palo Alto, open to all
Reviews current GENI status, Working Group meetings
Also discuss GPO solicitation, how to submit a proposal,
evaluation process & criteria, how much money, etc.
Travel grants to US academics for participant diversity
Subsequent Meetings, open to all who fit in the room
Held at regular 4-month periods
Held on / near university campuses (volunteers?)
All GPO-funded teams required to participate
Systematic, open review of each Working Group status
(all documents and prototypes / trials / etc.)
Also time for Working Groups to meet face-to-face
Results in prioritized list for next round of prototype funding
areas (priorities decided by NetSE Council and GPO)
AsianFI 2008 Summer School
21
Our Research Activities wrt PL/GENI
Network Virtualization Research Lab
Recruiting PostDocs/Ph.D. students
Private PlanetLab Infrastructure
GENI / VINI / PlanetLab
JGN2+ and JGN3
Federation of Test-Beds
Within Japan and Asia
With U.S. and Europe (OneLab2)
AKARI
New Generation Network Research
Virtualization as an Architecture
AsianFI 2008 Summer School
22
CORE: Private PlanetLab
Current:
CORE
Sapporo
• 10 sites, 52 servers
• Multi-Homed
Collaborative Overlay Research Environment
Overlay test-bed based on “Private PlanetLab”
Provision resources for mission critical services
Sapporo
Medical U.
Features we would like to have…
Custom hardware to optimize overlay forwarding
PoP/Core collocation (nodes “inside” network)
Custom hardware to optimize overlay forwarding
Tohoku U.
Federation (e.g. PlanetLab, OneLab)
Target overlay research
Sendai
Hiroshima U.
NII
Kanazawa
Kyutech
Not just on distributed system apps
More on network core architectures
Utilize both private & public environments
Nagano
Local v.s. Global / Provisioned v.s. Best-Effort
Tsukuba
Nagoya
Okayama
Kitakyushu
Tokyo
U. Tokyo
New Generation Perspectives to Overlay Network
OsakaKeihanna
Fukuoka
Kochi
- Testbed for prototype and evaluate
a new generation network design
- Evolutional nature of overlay network
to incorporate into the design
NICT Otemachi
Osaka U.
Kochi-tech
NICT
Koganei
AsianFI 2008 Summer School
23
PlanetLab Federation in Japan
Public
Production
Sapporo
Sapporo
Medical U.
• Operate public nodes
• Federation with PLC/PLE/PLA
• Boost PlanetLab communityHiroshima U.
• JGN2plus official service
• Private
Tohoku U. PlanetLab
• Promote PlanetLab in Japan
Sendai
NII
Kyutech
Kanazawa
Nagano
Tsukuba
USA
Nagoya
Chicago
Tokyo
Okayama
Kitakyushu
Osaka
Keihanna
U. Tokyo
Fukuoka
Kochi
NICT Otemachi
Osaka U.
Kochi-tech
Research
(Beta)
NICT Koganei
• GENI-like test-bed
• Network virtualization research
• Prototype of virtual routers
AsianFI 2008 Summer School
24
24
Net-Virt Lab Members
NAKAO Akihiro (Project Leader)
YAMAMOTO Shu (KDDI)
NAKAUCHI Kiyohide
OZAKI Ryota
NISHIDA Yuji
Joined from abroad…
Kurt TUTSCHK (Germany)
Aun HAIDER (New Zealand)
Yufeng WANG (China)
Richard Potter (U.S.)
Du Ping (China)
…inviting more researchers
3 more candidates (foreign)
Maoke Chen
JonWong Kim
MURATA Masayuki (Advisory)
ATA Shingo
OHZAHATA Satoshi
KAMEI Satoshi (NTT Lab)
KAWAHARA Ryoichi (NTT Lab)
Eng LUA (NTT Lab)
ENOMOTO Nobuyuki (NEC)
FUJINAMI Makoto (NEC)
NISHIKI Ken’ya (Hitachi)
SATOH Hiroki (Hitachi)
MATSUI Susumu (Hitachi)
SHUDO Kazuyuki (Utagoe/NICT)
KDDI Lab people
AsianFI 2008 Summer School
25
Research Lab Location
Hakusan, Tokyo (near Utokyo)
3-Floors (3F, 4F, 6F)
AsianFI 2008 Summer School
26
Our Research Activities wrt PL/GENI
Network Virtualization Research Lab
Recruiting PostDocs/Ph.D. students
Private PlanetLab Infrastructure
GENI / VINI / PlanetLab
JGN2+ and JGN3
Federation of Test-Beds
Within Japan and Asia
With U.S. and Europe (OneLab2)
AKARI
New Generation Network Research
AsianFI
2008 Summer School
Virtualization as an
Architecture
27