CSTalks - Named Data Networks - 9 Feb

24
Presenter: Mostafa Rezazad Supervisor: Profesor Y. C. Tay The third talk of CSTalk series

description

 

Transcript of CSTalks - Named Data Networks - 9 Feb

Page 1: CSTalks - Named Data Networks - 9 Feb

Presenter: Mostafa RezazadSupervisor: Profesor Y. C. Tay

The third talk of CSTalk series

Page 2: CSTalks - Named Data Networks - 9 Feb

Please ask as many questions as you can

24 pages 2/14/2011 2

Page 3: CSTalks - Named Data Networks - 9 Feb

Motivation Current Projects Introducing NDN Packet types and node structure Name Structure Routing Conclusion

24 pages 2/14/2011 3

Page 4: CSTalks - Named Data Networks - 9 Feb

Despite internets success, it is far from ideal. Why?

Now services and data are not the first class internet objects (the first class is the place)

Difficulties with mobility and multi homing Redundancy Security Middle Boxes problem (NATs, firewalls)

24 pages 2/14/2011 4

Page 5: CSTalks - Named Data Networks - 9 Feb

Resource sharing was the primary problem that networking aimed to solve during 60 and 70s.

The communication model is a conversation between exactly two machines.

Almost all the traffic on the Internet consists of TCP conversations between pairs of hosts

But what is the main goal of using network (Internet) todays? Accessing data or accessing machines!

24 pages 2/14/2011 5

Page 6: CSTalks - Named Data Networks - 9 Feb

Networks and consequently Internet were designed to accommodate communication between machines not applications and data.◦ DNS and IP two global namespaces are very rigid

When an application request a service or data it only cares about the content. However DNS based names for services force application to resolve service and data names down to IP address.

24 pages 2/14/2011 6

Page 7: CSTalks - Named Data Networks - 9 Feb

The increasing user demand for seamless communication on the move brings about new challenges that stress the current Internet, originally designed to support communications between fixed end-points.

24 pages 2/14/2011 7

Page 8: CSTalks - Named Data Networks - 9 Feb

Increase performance via Redundancy elimination◦ Mostly on application layer like web proxy caches Reduces the bandwidth usage of static content◦ Attempt to eliminate redundancy bellow the APP

layer They are not tied to a single application More redundant information can be removed

24 pages 2/14/2011 8

Page 9: CSTalks - Named Data Networks - 9 Feb

In order to have a secure communication we need to secure every part of the network, hosts, links, content and even clients.

Many threats comes to picture when the container is the subject of security.

24 pages 2/14/2011 9

Page 10: CSTalks - Named Data Networks - 9 Feb

They are designed to remedy the situation but they increase the complication

They are not generally part of the TCP/IP stack

24 pages 2/14/2011 10

Page 11: CSTalks - Named Data Networks - 9 Feb

There are four major funded projects each one worth up to $8 million over three years started from August 2010 (NSF site):◦ Named Data Networking

Principal Investigator: Lixia Zhang, UCLA◦ Collaborating Institutions: Colorado State

University, PARC, University of Arizona, University of Illinois/Urbana-Champaign, UC Irvine, University of Memphis, UC San Diego, Washington University, and Yale University

24 pages 2/14/2011 11

Page 12: CSTalks - Named Data Networks - 9 Feb

MobilityFirstPrincipal Investigator: Dipankar Raychaudhuri, Rutgers University/New BrunswickCollaborating Institutions: Duke University, Massachusetts Institute of Technology, University of Massachusetts/Amherst, University of Massachusetts/Lowell, University of Michigan, University of Nebraska/Lincoln, University of North Carolina/Chapel Hill

The project focuses on the tradeoffs between mobility and scalability and on opportunistic use of network resources to achieve effective communications among mobile endpoints.

24 pages 2/14/2011 12

Page 13: CSTalks - Named Data Networks - 9 Feb

NEBULAPrincipal Investigator: Jonathan Smith, University of PennsylvaniaCollaborating Institutions: Cornell University, Massachusetts Institute of Technology, Princeton University, Purdue University, Stanford University, Stevens Institute of Technology, University of California/Berkley, University of Delaware, University of Illinois/Urbana-Champaign, University of Texas, University of Washington

The project focuses on developing new trustworthy data, control and core networking approaches to support the emerging cloud computing model of always-available network services. This project addresses the technical challenges in creating a cloud-computing-centric architecture.

24 pages 2/14/2011 13

Page 14: CSTalks - Named Data Networks - 9 Feb

eXpressive Internet ArchitecturePrincipal Investigator: Peter Steenkiste, Carnegie Mellon UniversityCollaborating Institutions: Boston University, University of Wisconsin/Madison

XIA enables flexible context-dependent mechanisms for establishing trust between the communicating principals, bridging the gap between human and intrinsically secure identifiers.

24 pages 2/14/2011 14

Page 15: CSTalks - Named Data Networks - 9 Feb

They claim that the unified way to solve these problems is to replace where (or who) with What.

CCN or NDN has no notion of host at its lower level. A packet address names, not location.

NDN protocol stack is quite similar as the TCP/IP except Network layer and some refinement on layer 2.

One advantage of NDN is that it can be layered over anything, including IP itself.

24 pages 2/14/2011 15

Page 16: CSTalks - Named Data Networks - 9 Feb

Data satisfies an interest if the content name in the interest is a prefix of the content name in the data packet

Interest Interest

Interest

Interest

Interest Interest InterestInterest

DATA

DATA

DATA

DATA

DATA

Interest

Interest

DATA

24 pages 2/14/2011 16

Page 17: CSTalks - Named Data Networks - 9 Feb

A node has three data structure:◦ FIB: Forwarding Information Base It has a list of out going faces. It forwards interest

packets toward potential source(s)◦ Content Store: buffer memory Something like caches◦ PIT: Pending Interest Table Keeps track of upstream sending requests. Can be

used to down stream data

24 pages 2/14/2011 17

Page 18: CSTalks - Named Data Networks - 9 Feb

When an Interest packets appears on some interface, a longest match lookup is done on its content name. ◦ first the content will be searched from content store ◦ then over PIT table◦ And finally through the FIB table

24 pages 2/14/2011 18

Page 19: CSTalks - Named Data Networks - 9 Feb

Interest1 Interest1

Interest1

Interest1

Interest1 Interest1 Interest1Interest1

Interest1 (2)

PITInterest1 1

PITInterest1 1,2

Interest1 (2)

24 pages

DATA

DATA

DATA

DATA DATA

PIT

2/14/2011 19

Page 20: CSTalks - Named Data Networks - 9 Feb

Interest1

Interest1

Interest1

Interest1

Interest1

Interest1

/NUS/mostafa DATA

/NUS/mostafa 2

/NUS/mostafa DATA

FIB/NUS/mostafa 1

FIB

/NUS/mostafa 3,4FIB43

24 pages 2/14/2011 20

Page 21: CSTalks - Named Data Networks - 9 Feb

CCN name identify an information collection (not an information container)

Name hierarchy indicates membership The same information can have many names

(web like links) The hierarchical structure is used to do

longest match lookups (similar to IP prefix lookups) which helps guarantee log(n) state scaling for globally accessible data.

24 pages 2/14/2011 21

Page 22: CSTalks - Named Data Networks - 9 Feb

24 pages 2/14/2011 22

Page 23: CSTalks - Named Data Networks - 9 Feb

Existing link-state routing protocols can be used, unmodified, to construct a CCN FIB

24 pages 2/14/2011 23

Page 24: CSTalks - Named Data Networks - 9 Feb

Redundancy elimination No mobility restriction More secure Directly accessing the services without

knowing the place

24 pages 2/14/2011 24