Internet Architecture

Download Report

Transcript Internet Architecture

The End of Internet Architecture
Author: Timothy Roscoe
Presented by Gross, Zhaosheng Zhu
Outline
 Introduction to internet architecture
 Pressures for internet architecture
 Why architecture?
 GENI Platform and What we will do
 Some challenges
 Summary
2
Introduction to Internet
Architecture
 Internet Architecture does not act as a
useful guide
 Separates distributed systems from
networking
 Creates unnecessary problems for
distributed systems
3
Pressures for Internet Architecture
 Pressures from Within

Required Functionality
• QoS
• Security Against Attacks

Introduced Network Functionality
• Firewalls
• NAT boxes
4
Pressures for Internet Architecture
 Pressure from Above

Overlay technology
• Increases burden on underlying networks
• Increases difficulty to manage data flow
5
Pressures for Internet Architecture
 External Pressure

Gatewayed networks
• Phone Networks
• Sensor Network

Networks Proposed by Impasse paper
6
Why Architecture?
 Architectures fail to
Offer interoperability across diverse networks
 Allow for easier application coding
 Allow providers to compete
 Facilitate innovation

7
GENI platform: virtualization
 The increasing trend toward
programmability in networks needs
virtualization technology
 GENI Platform uses virtualization. The
hardware components can be sliced and
shared between different users
8
GENI makes things easy
 Experiments under GENI is not complex.
 Under GENI platform, You can use a set of
libraries and software management
services to control you experiment (acquire
resources and build ensembles).
9
GENI: invert the traditional layering of
applications and networks
 GENI substrate views experiments
embodying new networks architecture
as applications sharing the platform.
 But, in traditionally, it defines a
network architecture as a substrate
for applications.
10
Deploying a new network
architecture under GENI
1. Assemble a slice, that is, a set of
virtual servers, routers, links, radios,
sensors, etc.
2. Write and deploy the software
implementing the network
architecture to be used.
3. Write and deploy the software to
peer the network with the existing
internet in some way.
4. Write and deploy the newlyenabled services and applications.
 Why call
software
in step 2
network
architect
ure?
Just some
libraries!
11
However
 GENI does not define a new network architecture,
because:
 It is not creating a fictional boundary between two
disciplines.
 Structure of GENI leaves the question of end-toend connectivity an application defined issue.
 But writing applications are not becoming harder.
Remember we can reuse the libraries and services
provided by other parties.
12
What we suggest is:
 Take a application-
centric look
 Give users the
ability to create
virtual machines,
virtual routers and
virtual links and
users write
applications by
themselves
Benefits:
 Services use explicit
resources
 Complex cross-provider
bartering based on packet
measurement is not needed
at all!
 Carriers just care for
selling virtualized resources
which is a easier task
 More…
13
Why architecture?
Distributed system
Distributed system
Internet Arch.
Networking research
Networking research
14
Better, but still has challenges
Routing as a library
 Each application is effectively setting up its own
network. But until now there is little work on
simultaneous routing on many overlapping graphs
Discovery
 How do applications discover and bind to a set of
resources?
15
Some other challenges
Composition and federation
 As applications build their own networks,
how will they talk to each other?
Operations
 How will operators manage the substrate?
This is more like problem in distributed
systems.
16
Summary
 The concept of internet architecture has
been useful in history but now is an obstacle
for innovation.
 Removing it can make us move forward with
more basic problems such as how to build and
peer distributed applications.
 It is time to consider this.
17
Thank you !
18