L.GHAZZAI 1 Y.HADJADJ-AOUL 1 A.KSENTINI 1 G.BICHOT 2 S.GOUACHE 2 A.BELGHIT 3
                                           1 University of Rennes 1 2 Technicolor 3 University of Manouba




On a novel joint replicating and caching strategy for
              content centric networks
                    Portal                                                                Device  with  caching  


                                                                                                                                                                        ABSTRACT
                                                                                          support  (e.g.  Router,  
                                                                                              DSLAM,   )

                                                                                           STB  with  possible  
                                                                                            caching  support


                                                                                                                                The rise of popularity of video services has resulted in increased
                                                                                                                                   volumes of network traffic that, in turn, has created bottlenecks
                                                                                                                                   in the networks causing degradations of the perceived quality.
       CDN                                                                                                                         The CCN paradigm is considered as one of the most
     surrogate
                                                                                                                                   prominent solution to address such issue. However, the
                                                                                                                                   cascaded LRU caches introduced by CCN presents some
                                                                                                                                   limits. In this work, we first analyze such limits. Then, we
                                                                                                                                   propose a new caching and replication strategy to optimize
                                                                                                                                   resources utilization and to maximize the number of different
                                                                                                                                   chunks existing within the intra domain.
                                       Fig.1  Intra-­‐Domain  Architecture




                                                                                                                                                      Fig.2  Popularity  vs.  Number  of  Nodes
                                                                                                                                            Number of Nodes


 Content replication vs. caching in CCN                                                                                                18
                                                                                                                                       16
Early, in-network caching was proposed as a mean to get the                                                                            14
    contents closer to the end-users. With the shift towards                                                                           12
    content-centric networking (CCN), this logic is pushed further.
                                                                                                                                       10
    CCN introduces two distinct techniques: contents caching and
                                                                                                                                        8
    replication. However, one should consider the mutual impact
    existing between these techniques. Indeed, the benefits of                                                                          6

    contents replication can be completely cancelled with a                                                                             4
    bad caching technique (see Fig. 3).                                                                                                 2
                                                                                                                                                                                                             Popularity

                                                                                                                                        0
                                                                                                                                             0                5E-­‐11    1E-­‐10   1,5E-­‐10   2E-­‐10   2,5E-­‐10

                               Fig.3  Popularity  vs.  Delay
     0,8
                                                                                                                                            Limits of existing approaches
           Delay



     0,7
     0,6
                                                                                                                                CCNs allow popular content to be present in many nodes to make
     0,5                                                                                                                          the content closer to the end users (see Fig. 2). However, the
     0,4                                                                                                                          use of LRU as a caching strategy deceases the duration of
     0,3                                                                                                                          the contents presence in caches.
     0,2                                                                                                                        Some changes should be introduced to the classical CCN
     0,1                                                                                                                           architecture by focusing on: (i) reducing the amount of
      0                                                                                                                            replica in the intra-domain; (ii) storing as many various data
                                                                                                                   Popularity
           0                 5E-­‐11            1E-­‐10             1,5E-­‐10   2E-­‐10                2,5E-­‐10                   as possible.



   A Combined caching and replication                                                                                               Each CCN node, in the path to the destination, stores the chunk
                                                                                                                                    using this probability.
               technique
                                                                                                                                    The proposed caching technique, combines the benefits of
   Initially affect to each piece of data (i.e. chunk), to be                                                                       Least Recently Used (LRU) and Least Frequently Used (LFU)
   transmitted, a nonzero storage probability depending notably on                                                                  solutions.
   the chunk s popularity P.
                                                                                                                                    If a CCN node decides to store a particular chunk, it puts
                      P0 = min [ max (( i) , 0) , 1]
   P=                                                                                                                               its storage probability to zero (or reduce the probability) to
                      PK = min[max(PK-1 + S /N , 0) , 1]
where i : popularity of the chunk i. S = 1 if the chunk have been                                                                   avoid multiple duplication of the same content. Otherwise,
  stocked, 0 else.                                                                                                                  the probability is increased.


                   Acknowledgement                                       The work is partially supported by the national French project ANR VERSO ViPeer.

CCNxCon2012: Poster Session:On a Novel Joint Replicating and Caching Strategy for Content-Centric Networks

  • 1.
    L.GHAZZAI 1 Y.HADJADJ-AOUL1 A.KSENTINI 1 G.BICHOT 2 S.GOUACHE 2 A.BELGHIT 3 1 University of Rennes 1 2 Technicolor 3 University of Manouba On a novel joint replicating and caching strategy for content centric networks Portal Device  with  caching   ABSTRACT support  (e.g.  Router,   DSLAM,   ) STB  with  possible   caching  support The rise of popularity of video services has resulted in increased volumes of network traffic that, in turn, has created bottlenecks in the networks causing degradations of the perceived quality. CDN   The CCN paradigm is considered as one of the most surrogate prominent solution to address such issue. However, the cascaded LRU caches introduced by CCN presents some limits. In this work, we first analyze such limits. Then, we propose a new caching and replication strategy to optimize resources utilization and to maximize the number of different chunks existing within the intra domain. Fig.1  Intra-­‐Domain  Architecture Fig.2  Popularity  vs.  Number  of  Nodes Number of Nodes Content replication vs. caching in CCN 18 16 Early, in-network caching was proposed as a mean to get the 14 contents closer to the end-users. With the shift towards 12 content-centric networking (CCN), this logic is pushed further. 10 CCN introduces two distinct techniques: contents caching and 8 replication. However, one should consider the mutual impact existing between these techniques. Indeed, the benefits of 6 contents replication can be completely cancelled with a 4 bad caching technique (see Fig. 3). 2 Popularity 0 0 5E-­‐11 1E-­‐10 1,5E-­‐10 2E-­‐10 2,5E-­‐10 Fig.3  Popularity  vs.  Delay 0,8 Limits of existing approaches Delay 0,7 0,6 CCNs allow popular content to be present in many nodes to make 0,5 the content closer to the end users (see Fig. 2). However, the 0,4 use of LRU as a caching strategy deceases the duration of 0,3 the contents presence in caches. 0,2 Some changes should be introduced to the classical CCN 0,1 architecture by focusing on: (i) reducing the amount of 0 replica in the intra-domain; (ii) storing as many various data Popularity 0 5E-­‐11 1E-­‐10 1,5E-­‐10 2E-­‐10 2,5E-­‐10 as possible. A Combined caching and replication Each CCN node, in the path to the destination, stores the chunk using this probability. technique The proposed caching technique, combines the benefits of Initially affect to each piece of data (i.e. chunk), to be Least Recently Used (LRU) and Least Frequently Used (LFU) transmitted, a nonzero storage probability depending notably on solutions. the chunk s popularity P. If a CCN node decides to store a particular chunk, it puts P0 = min [ max (( i) , 0) , 1] P= its storage probability to zero (or reduce the probability) to PK = min[max(PK-1 + S /N , 0) , 1] where i : popularity of the chunk i. S = 1 if the chunk have been avoid multiple duplication of the same content. Otherwise, stocked, 0 else. the probability is increased. Acknowledgement The work is partially supported by the national French project ANR VERSO ViPeer.