page 1 content centric network: caching wang yu katto lab. dec.6 2012

13
Page 1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Upload: philomena-freeman

Post on 03-Jan-2016

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 1

Content Centric Network: Caching

WANG Yu KATTO Lab.Dec.6 2012

Page 2: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 2

CCN (content centric network)

Content centric network is a new communication architecture that rethinks the Internet communication model, and centers it around content dissemination and retrieval.

Main Characteristic

1. Request driven communication

2. Hierarchical name based routing

3. Content store

Page 3: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 3

Content Router or Caching router (CR) New generation of routers Cache packets of content and reuses those that are still in the cache when su

bsequently requested Maximize the probability of sharing,minimizes upstream bandwidth demand an

d downstream latency. [1]

CCN caching strategy proposed by Van Jacobson Least Recently Used(LRU) or least Frequently Used(LFU) [1]

[1] Networking Named Content, Van Jacobson, Diana K. Smetters, James D. Thornton, Michael F. Plass, Nicholas H. Briggs, Rebecca L. Braynard, Palo Alto Research Center, Palo Alto, CA, USA

Page 4: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 4

what's proposed in CCN caching?

LRU

Discards the least recently used items when the cache is full

LFU

Discards the lowest frequently used items when the cache is full

Trade-off between recency and frequency

Combine LRU and LFU, assign them with different weights.

Page 5: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 5

Problem with LRU ?

In [2], the author models and evaluates the caching feature of CCN caching strategy LRU using NS2. And it concludes there is a clear network-wide performance gain for popular content, but this doesn't work for unpopular content, and the performance heavily depends on the storage size of router cache too. So simply LRU are not good enough.

[2] Modelling and Evaluation of CCN-caching Trees, Ioannis Psaras, Richard G. Clegg, Raul Landa, Wei K. Chai, and George Pavlou, Department of Electronic &Electrical Engineering University College London

Page 6: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 6

Better CCN caching policies?

Cooperative In-Network Caching

weighted caching chunk in CCN

Page 7: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 7

Cooperative Caching

chunk{11,..,20}

chunk{11,..,20}

chunk{11,..,20}

chunk{0,3,6,9,12,15,18}

chunk{2,5,8,11,14,17,20}

chunk{1,4,7,10,13,16,19}

interest response of CR response of Server

LRU caching :

Cooperative Caching :

chuck 5?

chuck 15?

chuck 5?

chuck 15?

chuck 5?

chuck 5?

chuck 5?

chuck 5?

chuc

k 1

5?

Page 8: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 8

New tables added in CCN [3]

1. Collaborative Router Table(CRT)

2. Collaborative Content Store(CCS)

How to store caches cooperatively?

Cooperative Caching

[3] Time-shifted TV in Content Centric Networks: the Case for Cooperative In-Network Caching, Zhe Li, Gwendal Simon, Institut Telecom- Telecom Bretagne, France, IEEE Communication Society 2011

client1

client2

chuck 21? chuck 21?chuck 21? chuck 21?

chuck 21chuck 21

21mod3=0≠2, check CRT table, so send chunk 21 to r0 to be cached

label identifier interface

0 r0 2

1 r1 6

chuck 21 arrived, check its CRT, then cache chunk 21 in its Content Store

CRT of r2

Prefix (requested content)

requesting face

chunk 21 6

... ...

PIT table of r2

chuck 21chuck 21

chuck 212

6

content name label interface

chunk 21 0 2

CCS table of r2

Page 9: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 9

Article [3] uses simulation to compare Basic LRU with cooperative Cache and it finds:

Cooperative Cache has better caching diversity

ISP-friendliness: Under cooperative cache strategy the number of times which each server located is accessed has minimized

Besides cooperative caching strategy, when a cache router's storage is full, should we just remove the least recently used cache chunk (LRU)?

Cooperative Caching

Page 10: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 10

Storage Management Polices

Weighted storage of CCN caching chunk

But I am still considering and researching how it actually works.

Previous work mainly focus on content-level replacement policies based on different criteria as recency, frequency,size, QoS. But nowadays,the Internet are more and more diverse. For example: wireless phones, web-enabled vehicles. We may need to prioritizly cache some chunks because of QoE or /and pricing motivations instead of just using LRU, LFU. So we may need more flexible caching in each content router. I think using weighted caching chunk in CCN may solve such problem.

Page 11: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 11

Problem still existed?

most of the newly proposed CCN caching algorithms are conducted using simulators rather than in realistic network condition, and some assumptions are simplified, so it may lack accuracy.

Page 12: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 12

Where should we go?

Next step, I will try to think about weighted caching in CCN

And also try to simulate some CCN algorithms in relatively realistic network condition.

Page 13: Page  1 Content Centric Network: Caching WANG Yu KATTO Lab. Dec.6 2012

Page 13

Thank you