Economic Traffic Management (ETM) Mechanisms – Selected View


Published on

Published in: Technology, Business
  • Be the first to comment

Economic Traffic Management (ETM) Mechanisms – Selected View

  1. 1. © 2010 The SmoothIT Consortium 1 Economic Traffic Management (ETM) Mechanisms – Selected View SSimple Economicimple Economic MManagement Approachesanagement Approaches ooff OOverlayverlay TTraffic inraffic in HHeterogeneouseterogeneous IInternetnternet TTopologiesopologies European Seventh Framework STREP FP7-2007-ICT-216259European Seventh Framework STREP FP7-2007-ICT-216259 UZH, DOCOMO, TUD, AUEB, PrimeTel, AGH, ICOM, UniWue, TIDUZH, DOCOMO, TUD, AUEB, PrimeTel, AGH, ICOM, UniWue, TID Fabio Hecht, UZHFabio Hecht, UZH (on behalf of SmoothIT)(on behalf of SmoothIT) October 20, 2010October 20, 2010 Brussels, BelgiumBrussels, Belgium
  2. 2. © 2010 The SmoothIT Consortium 2 Basics and Motivation  Use of economic mechanisms for controlling, managing network traffic of overlays at early stages:  SmoothIT results show already that such mechanisms do have the important property of scalability and effectiveness! – Situation-dependent ETMs lead to a more efficient network operation – ETMs generate a higher value (QoE) for its customers.  In managing the traffic created and routed through their networks, today’s ISPs are offered by SmoothIT methodologies suitable for modern traffic/service profiles – E.g., peer-to-peer traffic is treated according to ETM approaches. – Applicable to traffic of different P2P applications Economic Traffic Management (ETM)
  3. 3. © 2010 The SmoothIT Consortium 3 Triple Win  All ETM mechanisms aim at achieving TripleWin – Overlay traffic is optimized beneficially for all three stakeholders: ISPs, overlay providers, and users  Incentives for operators (i.e. ISPs) – Monetary: reduce overlay traffic and inter-domain traffic – Traffic management: less congested links, better performance – Reputation: keep customers, distinguish from other operators  Incentives for overlay providers – Performance: Active role in traffic mgmt increases service quality – Reputation: increased user base due to better performing services  Incentives for users – Performance: Increased service quality, e.g., reliability, RTT, BW – Monetary: lower price for network access
  4. 4. © 2010 The SmoothIT Consortium 4 ETM Classification and Synergies  High potential for synergies  Progressive development of approaches possible  Incremental enhancement of architecture possible
  5. 5. © 2010 The SmoothIT Consortium 5 ETM Mechanisms Ported to Trials 1. BGP-Based Locality Promotion (BGP-Loc): ISP provides proximity-related recommendation to overlay applications, which is used to optimize traffic. 2. Insertion of ISP-Owned Peers (IoP): Resourceful entity enhancing both traffic locality and QoE within an ISP. 3. Promotion of Highly Active Peers (HAP): Boosts peers which contribute the most in order to achieve similar results as IoPs with little investment in infrastructure.
  6. 6. © 2010 The SmoothIT Consortium 6 SmoothIT Information Service (SIS)  Deployment of SIS components in the ISPs’ network – To convey information between overlay and underlay  Client-Server architecture  Overlay applications interact with SIS in order to select “better” peers, e.g. local peers, IoPs or HAPs. – Reducing ISPs costs and improving QoE of users SIS P P P P P P P SIS ISP A ISP B SIS protocol P Peer / Overlay appl.
  7. 7. © 2010 The SmoothIT Consortium 7 SmoothIT Information Service (SIS) IoP ETMS Underl ay Overla y SIS H A P (Aggregate) underlay metrics Peers’ (abstracted) overlay status Popular swarms Best peers QoS enhancements QoE improvements Cost reduction Revenue increase SIS is the core of the ETM System (ETMS), which can lead to TripleWin.
  8. 8. © 2010 The SmoothIT Consortium 8 Simulated Network Topology Hub AS 1 Initial seeder SIS Transit AS Hub AS 2 Stub AS 1…10 Stub AS 11…20 Peering link Transit link Inter-AS link
  9. 9. © 2010 The SmoothIT Consortium 9 BGP-Loc  At application: task is to discover neighbors to download from and select neighbors to upload to – “Unchoked” neighbors receive data – “Choked” neighbors do not receive data.  Used BitTorrent (file sharing) and Tribler (video streaming, based on BitTorrent) – Regular BitTorrent (regBT): • TFT slots: upload to peers that have provided the most • optimistic unchoking slot: upload to a random peer – BNS: Biased Neighbor Selection • Peers prefer to download from local peers from SIS – BOU: Biased Optimistic Unchoking • Peers prefer to upload to local peers from SIS
  10. 10. © 2010 The SmoothIT Consortium 10 Access inter-AS 0 2 4 6 8 10 12 14 16 Bottleneck Type DownloadTimes(min) regBT BOU BNS BNS&BOU BGP-Loc: Simulation Results Homogeneous Scenario Mean traffic (Mbyte/s) Mean download times (min) Access inter-AS 0 10 20 30 40 50 Bottleneck Type Bandwidth(MB/s) Intra-AS Peering Links Transit Links regBT BOU BNS BNS&BOU BitTorrent: Bottleneck Types Inter-AS Inter-AS Inter-AS Inter-AS
  11. 11. © 2010 The SmoothIT Consortium 11 Mean traffic (Mbyte/s) Mean stalling times (s) Tribler: Bottleneck Types Access Core 0 500 1000 1500 2000 2500 Bottleneck Type StallTimes(sec) regBT BOU BNS BNS&BOU Access Core 0 5 10 15 20 25 30 35 40 45 50 Bottleneck Type Bandwidth(MB/s) Intra-AS Peering Links Transit Links regBT BOU BNS BNS&BOU Inter-AS Inter-AS Inter-AS Inter-AS BGP-Loc: Simulation Results Homogeneous Scenario (2)
  12. 12. © 2010 The SmoothIT Consortium 12 BGP-Loc: Simulation Results Heterogeneous Scenario Mean upload traffic for different ASes with different populations (Mbyte/s) Mean download times (min) Heterogeneous peer distribution 0 5 10 15 20 8 9 10 11 12 13 14 15 AS ID DownloadTimes(min) Ref BU BNS BNSBU 0 5 10 15 20 0 0.5 1 1.5 2 2.5 3 3.5 AS ID UploadBandwidth(MB/s) Ref BU BNS BNSBU
  13. 13. © 2010 The SmoothIT Consortium 13 BGP-Loc: Evaluation Summary  Bottleneck: access links  Win–No-lose – Download times remain unaffected – Savings in inter-AS bandwidth are achieved  Bottleneck: inter-domain links  Win–Win – Still some inter-AS bandwidth can be saved – Download/stalling times can be improved significantly  The efficiency of locality promotion is higher in ASes having larger fractions of the swarm  The performance improvement further increases with the fraction of locality-promoting peers
  14. 14. © 2010 The SmoothIT Consortium 14 Insertion of ISP-owned Peers (IoP)  Motivation: Sole locality may not improve peers’ performance. We can exploit overlay functioning to localize traffic and achieve Win-Win  Approach: Insert an ISP-owned peer (overlay entity) provisioned with higher access capacity  Impact: Improvement of peers’ performance and reduction of inbound traffic  Innovation: Transparency, no interception required. Variety of policies
  15. 15. © 2010 The SmoothIT Consortium 15 Promotion of Highly Active Peers (HAP)  Motivation: Increase the access capacity of very active regular peers instead of inserting ISP-owned entities  Approach: Exploitation of ISP’s NGN capabilities to change the access profile of certain users  Impact: Give peers the incentive to serve as seeds. Localize traffic while improving peers’ performance.  Innovation: Fully innovative mechanism, with NGN. Extra resources directly given to peers.
  16. 16. © 2010 The SmoothIT Consortium 16 Summary and Conclusions  Dedicated management of overlay traffic is necessary – Due to smoothing large amounts of overlay traffic – Due to the minimization of high(er) costs for ISPs  SmoothIT architectural design and impl. completed – SmoothIT Information Service (SIS) • Deployed in networks of ISPs • Provides information to overlay applications • Optimizes traffic and achieves the Triple Win situation  Many sets of simulative evaluations in place  SmoothIT participates at IETF in ALTO (3 drafts)  Trials with selected ETM mechanisms are running currently in a real network
  17. 17. © 2010 The SmoothIT Consortium 17 Thank you for your attention! Thanks to all SmoothIT’s project partners: UZH, DOCOMO, TUD, AUEB, PrimeTel, AGH, ICOM, UniWue, TID