Next Generation Network Architectures...6 Angry Birds 4.15 0.15 10 Angry Birds Rio 4.03 0.14 10...

Next Generation Network Architectures

Srinivasan Seshan!

Living Analy+cs

•  Rich data collec,on à real-‐,me data analy,cs à automated applica,on feedback à rich data collec,on…

•  Key networking/distributed systems challenges –  Scaling data collec,on –  Privacy/anonymity –  Localiza,on – More efficient networks

•  Leverage a range of work to address these challenges

2

Outline

•  Other Projects – Mobile Adver,sing –  Femto-‐cell Infrastructure Op,miza,on –  Localiza,on –  BeLer Mul,media URLs –  Spectrum Management

•  eXpressive Internet Architecture (XIA) •  Reliable Video and Real-‐,me Data

3

Free Mobile Apps and Mobile Advertising

80% of Mobile Apps Downloaded are Ad-

Supported (Free)

Significant Game Developer Revenues from Advertisements

($87M per month)

Telecom Carriers are Moving to Metered

Data Plans

Traffic “Cost” of Advertisement Content May be Substantial

4

How Substantial is the Cost? Rank Name Rate of ads

(kbps) App Data Rate (kbps)

6 Angry Birds 4.15 0.15 10 Angry Birds Rio 4.03 0.14 10 Hanging with Friends 1.76 5.86

11 Talking Tom 2 1.71 0.02 14 Words with Friends 0.89 2.44 19 Dictionary.com 3.69 9.74

30 mins of Angry Birds: 36 MB of traffic/month à 56c/month on Verizon’s 2 GB/$30 plan

HTML 5 based “rich” ads are typically 5-20 times larger in size

5

Redesigning Advertisement Delivery

•  Aggressive prefetching using inexpensive connec,vity –  Home/Office WiFi; Femtocells; Underused Macrocells

•  Challenges: –  Predic,ng context –  Preven,ng click-‐fraud –  Privacy –  Alternate charging/accoun,ng models –  Tight integra,on between mobile OS, mobile applica,on, telco, and adver,sing networks

6

Outline



10

XIA: An Architecture for a Trustworthy and Evolvable Internet

Peter Steenkiste, Dave Andersen, David Eckhardt, Sara Kiesler, Jon Peha, Adrian Perrig, Srini Seshan, Marvin Sirbu, Hui Zhang Aditya Akella John Byers

11

IP: Narrow Waist of the Internet

IP

Applications

Technology

Innovation both above and below IP

But what about IP?

12

Proposed -Centric Networking

•  Content: Named Data Networking •  Mobility: MobilityFirst •  Cloud: Nebula

13

Can we support heterogeneous communication types on a single Internet architecture?

Problem: Focusing on one communication type may hinder using other communication types, as occurred to IP

Future -Centric Networking

•  Service, content, mobility, and cloud did not receive much aLen,on before

•  Yet more networking styles may be useful in the future –  E.g., DTN, wide-‐area mul,cast, …?

14

Can we support future communication types without redesigning the Internet architecture?

Problem: Introducing additional communication types to the existing network can be very challenging

Legacy Router May Prevent Innovation

15

Can we allow using a new communication type even when the network is yet to natively support it?

Problem: Using a new communication type may require every legacy router in the network to be upgraded

“I got a computer with Awesome-Networking

announced at Sigcomm 2022! Can I use it right now?”

Internet

“Ouch, we just replaced all of our routers built in 2012. Can you wait for another

10 years for new routers?”

XIA’s Goals and Design Pillars

16

Support mul,ple communica,on

types (heterogeneity)

Support future communica,on

types (evolu,on)

Allow using new communica,on types at any point

(incremental deployment)

“Principal types” “Fallbacks”

17

Principal Types

Define your own communication model

Principals

18

128.2.10.162

Current Internet XIA

IP address

Host 0xF63C7A4…

Principal type

Type-specific identifier

Service 0x8A37037…

Content 0x47BF217…

Future …

Hash of host’s public key

Hash of content

Hash of service’s public key

Principal Type-Specific Semantics

19

Contact a host

Use a service

Retrieve content

Host 0xF63C7A4…

Service 0x8A37037…

Content 0x47BF217…

Principal Type-Specific Processing

20

XIA router

Host-specific processing

Common processing

Service-specific processing

Content-specific processing

…

Input Output

•  Type-specific processing examples •  Service: load balancing or service migration •  Content: content caching

Routers with Different Capabilities

•  Routers are not required to support every principal type –  The only requirement: Host-‐based communica,on

21

Host

Common

Host-‐only router

Host

Common Service

Service-‐enabled router

Host

Common

Content-‐enabled router

Content

Using Principal Types that are Not Understood by Legacy Routers?

22

Legacy router without

content support

Want to communicate using content principals

Content-enabled router

Content-enabled router

23

Fallbacks

Tomorrow’s communication types… today!

Fallbacks: Alternative Ways for Routers to Fulfill Intent of Packet

24

Content

Intent: Retrieve

Fallback: Contact ,

who understands request

What the network does:

•  With content-‐enabled routers, use for rou,ng

•  Otherwise, use for rou,ng (always succeeds)

Content

Host

Host

Content

25

DAG-Based Address

Your address is more than a number

DAG (Direct Acyclic Graph)-Based Addressing Enables Fallbacks

26

Intent Packet sender Rou,ng choice

Another rou,ng choice (with lower priority)

This host knows how to handle content request

Fallback

Content

Host

Service Binding with DAG

27

Service Web service

Ini,al contact to a service

Service Web service Host Server #57

When a par,cular host should serve subsequent service requests

“Late binding”

Incremental Deployment of XIA on IPv4 Network

•  4ID: IPv4 address as an XID –  IPv4 tunneling between XIA network islands –  Typically used as a fallback

4IDS

HIDS ADS

28

Represents IPv4 address of ADS

4ID in Action (1) Partially Deployed XIA Networks

29

XIA Network C XIA Network S IPv4 Network

4IDS

ADS

Entering IPv4 network: Encapsulate XIA packet with IP

header

Entering XIA network: Remove IP header for na,ve XIA

packet processing

Works for arbitrary pairs of XIA networks

4ID in Action (2) Fully Deployed XIA Networks

30

XIA Network C XIA Network S XIA Network C

4IDS

ADS

Use na,ve XIA forwarding and ignore fallback

Seamless incremental deployment of XIA

31

Can We Forward DAGs Rapidly?

Expressive ≠ Expensive

XIA Software Router’s High Forwarding Throughput

32

Click-‐based implementa,on on commodity hardware 351 K table entries based on a Route Views snapshot

≤26% slowdown for small packets with 3 fallbacks

•  Support for evolvable internetworking –  Principal types à heterogeneity –  Fallbacks & DAG-‐based addressing à incremental deployment

•  Prototype: www.xia.cs.cmu.edu

XIA: Enabling Evolution by eXpression

33

Something New

Something Old

Outline



34

Redundancy Elimination (RE)

•  10+ vendors Cisco, Riverbed, BlueCoat •  Corporate network, inter-‐data center deployment

Enterprise

Packet-Cache Packet-Cache

Access link Internet

35

•  Product: WAN op,mizers (10+ vendors) –  Cisco, Riverbed, Juniper, Blue Coat Systems –  E.g., Cisco deployed RE on 200+ remote offices. –  Corporate networks

•  Riverbed: 50+ corporate customers, datacenter deployments

Deployment of content-aware networks

36

Main office

Branch

WAN op,mizer

WAN op,mizer

VPN (“Virtual wire”)

Isola,on from Cross traffic

Can RE help real-time applications?

Time critical inter-data center communication [Maelstrom]

Soft-realtime intra-data center communication [DCTCP, D3]

Real-time streams: FaceTime, Skype, on-line games.

Protecting data loss in time-critical communication is important, but challenging because of the time constraint (~150ms)

37

Loss protection today: Redundancy-based recovery Forward Error Correction

Original packets (k)

Bandwidth for robustness

Redundant packets (n-‐k)

•  FEC couples delay with redundancy •  Small batch size makes FEC more suscep,ble to bursty loss •  Difficult to tune parameters (n and k) [TIP2001,INFOCOM2010]

Amount of redundancy 20%~50% in Skype video[Mul,media’09]

38

Delay

RE Network

Redundant Packet Transmission

40

FEC Redundancy Elimina,on Router re

dund

ant

RE Networks

Redundant Packet Transmission

41

Redundant Transmission Ques,ons/Challenges •  How do we make sure we retain the robustness benefits? •  How much redundancy is needed? How does it compare with FEC? •  Is this safe to use?

•  Introduce redundancy in a way that the network understands

Analytical Comparison with FEC

0.00

0.10

0.20

0.30

0.40

0.50

1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1

Frac,o

n of Overhead �

End-to-end data loss rate (%)�

RPT(3) RPT(2)

FEC(10,8)

FEC(10,7)

FEC(10,9)

FEC(10,6)

FEC(10,5)

RPT(4)

2% random loss. = 0.02

42

Naive 2% data loss 0 overhead

Naive

Batch size (n=10)

Delay

…

Original pkts (k=8)

FEC(n=10,k=8)

Redundancy (r=3) Delay RPT(3)

Coded redundancy

E2E Performance: Video Quality

43

Naïve UDP

RPT(3) Overhead ~6%

FEC(10,9) Overhead ~10%

1.8dB ~ 3dB difference in quality

30

31

32

33

34

35

36

37

38 Encoded video at sender A

vera

ge P

SN

R (d

B)


44

RPT FEC

(Before loss)

30

31

32

33

34

35

36

37

38 Encoded video at sender Received video A

vera

ge P

SN

R (

dB)


45 RPT FEC

Packet loss rate ~2%

•  RPT flows get priori,zed à need new conges,on control techniques

0.8 0.85 0.9 0.95 1

Sender

Receiver (Non-‐RPT)

Receiver (RPT)

Original Redundancy

0

Bandwidth use (Mbps)

0

9% loss

Problem: Impact on Other Traffic

46

Throughput reduc,on: 2%

(Before loss)

(A|er loss)

Packet loss rate : 9%.

RPT Flows

Loss

Other Flows

Loss

(A|er loss)

RPT Summary

•  Key Idea of RPT: Don’t hide, expose redundancy in content-‐aware networks! – Need to redesign end-‐points when you change the network

•  Key Features – High robustness, low overhead à user performance

–  Ease of use: parameter selec,on, per-‐packet redundancy/delay control

47

Living Analy+cs

•  Rich data collec,on à real-‐,me data analy,cs à automated applica,on feedback à rich data collec,on…

•  Key networking/distributed systems challenges –  Scaling data collec,on –  Privacy/anonymity –  Localiza,on – More efficient networks

•  Leverage a range of work to address these challenges

48

Next Generation Network Architectures...6 Angry Birds 4.15 0.15 10 Angry Birds Rio 4.03 0.14 10...

Documents

Transcript of Next Generation Network Architectures...6 Angry Birds 4.15 0.15 10 Angry Birds Rio 4.03 0.14 10...