RAIDER: RAIDER: Responsive Responsive Architecture for Architecture for Inter Inter-Domain Economics and -Domain Economics and Routing Routing Nirmala Shenoy Nirmala Shenoy, Rochester Institute of T , Rochester Institute of Technology echnology Murat Murat Yuksel, University of Nevada Reno uksel, University of Nevada Reno Aparna Gupta, Koushik Kar Aparna Gupta, Koushik Kar, Rensselaer Polytechnic Inst , Rensselaer Polytechnic Inst Vi Victor Perotti, Rochester Institute of Technology Manish Karir Manish Karir, Merit Networks , Merit Networks National Science Foundation funded.. National Science Foundation funded..
Outline Outline Goals of RAIDER –> Future Internet Goals of RAIDER –> Future Internet Components of RAIDER Components of RAIDER Networking Component Networking Component Floating Cloud T Floating Cloud Tiered Internetworking Model iered Internetworking Model Service Provisioning Component Service Provisioning Component Contract-Based Inter Domain Routing Contract-Based Inter Domain Routing Economic Component Economic Component Inter Inter-Domain Economics and Risk Management -Domain Economics and Risk Management Summary Summary Position paper – Individual results Position paper – Individual results
Goals of RAIDER Goals of RAIDER An internetworking architecture An internetworking architecture Highly Flexible and Scalable Highly Flexible and Scalable Te Technically and Economically - Respond to future needs of Network Respond to future needs of Network Users and Providers Users and Providers
RAIDER – T RAIDER – Technical and echnical and Economic Components Economic Components omics ent emen onom and Risk Managem Floating Cloud T Floating Cloud Tiered iered omain Econ Internetworking Internetwork ing Model Model -Dom er-D Contract Switching Contract Switching Inter
Floating Cloud T Floating Cloud Tiered (FCT) iered (FCT) Internetwork Internetworking Model ing Model Te Technically Responsive Architecture Modularity Modularity Granularity Granularity Structure to leverage Structure to leverage High Interconnections High Interconnections Address mechanism Address mechanism > Implement structure, avoid logical > Implement structure, avoid logical address based routing address based routing
FCT Internetworking Model FCT Internetworking Model Building Blocks Building Blocks Network Clouds – ISPs, POPs, Network Clouds – ISPs, POPs, Backbone routers…… Backbone routers…… Nested Clouds Nested Clouds Ti Tiers – Global Level – ISPs, AS – backbone, distribution, access… backbone, distribution, access… Nested T Nested Tiers iers
FCT – FCT – Applied – ISP Applied – ISP Level Level
Nested Clouds, T Nested Clouds, Tiers, iers, Addresses Addresses Inside a POP 1.1 Backbone Tier 1 2.1:2 2.1:1 Distribution AS D Distribution ISP Tier 2 C 3.1:1:2 3.1:1:1 Tier 3 Access Access Nested Nested Address 1.1{3.1:1:2} Address 1.1{3.1:1:2} Nested T Nested Tiers iers
Contract Contract Switching Switching
10 Inter Inter-domain Struggles… -domain Struggles… When crossing domains, all bets are off.. When crossing domains, all bets are off.. End-to-end reliability or performance-criticality End-to-end reliability or performance-criticality requires requires assurance of single-domain performance, i.e., “ assurance of single-domain performance, i.e., “contract”s contract”s efficient concatenation of single-domain contracts efficient concatenation of single-domain contracts Inter Inter-domain routing needs to be aware of economic -domain routing needs to be aware of economic semantics semantics contract routing + risk management contract routing + risk management We We address translation of these struggles to architectural problems architectural problems
Contract-switching: ISP A Paradigm Shift… B ISP A ISP e2e e2 e circuits C Circuit-s -switching ISP B rou outable e ISP A datagrams datagrams ISP Packet Pa et-s -switching C ISP B con ontracts ISP A over ov erlaid on on rou outable e ISP Con ontract-s -switching datagrams datagrams C 11
Basic Building Block: Intra- domain Dynamic Contracts Contract components Contract components performance performance component, e.g., capacity component, e.g., capacity financial financial component, e.g., price component, e.g., price time time component, e.g., term component, e.g., term Stations of the Stations of the provider computing provider computing Edge and advertising local and advertising local Router Edge prices for edge-to- prices for edge-to- Router edge contracts. edge contracts. Edge Router Network Core Customers accessed only Edge by contracts Router Edge Edge Router Router 12
Contract Link An ISP An ISP is abstracted as a is abstracted as a set of “contract links” set of “contract links” Contract link Contract link: : an an advertisable advertisable contract contract between peering/edge points between peering/edge points i and j of an ISP and j of an ISP with flexibility of with flexibility of advertising different prices advertising different prices for edge-to-edge (g2g) intra- for edge-to-edge (g2g) intra- domain paths domain paths capability of of managing value e fl flow ows at a fi finer er granularity than poi oint-t -to- o- anywher ere e dea eals 13
How to Achieve e2e QoS? Contract Routing: Contract Routing: Compose e2e inter Compose e2e inter-domain “contract paths” over available -domain “contract paths” over available contract links satisfying the QoS contract links satisfying the QoS requirements requirements Calculate the contract paths by shortest-path Calculate the contract paths by shortest-path algos algos with with metrics customized metrics customized w. w.r.t. contract . contract QoS QoS metrics metrics Tw Two ways: link-state contract routing link-state contract routing at macro time-scales at macro time-scales path-vector contract routing path-vector contract routing at micro time-scales at micro time-scales Monitor and verify that each ISP Monitor and verify that each ISP involved in an e2e involved in an e2e contract path is doing the job contract path is doing the job Punish the ISPs not doing their job, e.g. as a money- Punish the ISPs not doing their job, e.g. as a money- back guarantee to the others involved in the e2e back guarantee to the others involved in the e2e contract path contract path 14
Path-V Path-Vector Contract Routing: ector Contract Routing: Micro-level, On-demand, Reactive Micro-level, On-demand, Reactive [C-B-A, 5-4-2-1, 20Mb/s, 30mins, $7.3+$3] [C-B, 5-4-2, 20Mb/s, 45mins, $6+$5] [C, 5-4, 30Mb/s, ISP h t a p t n e m e c n u B 45mins, $9] o n 2 n a ISP 1 4 Provider Provider A User X initiates… initiates… 3 path ISP announcement ISP ISP C wants to C wants to C advertise advertise 5 availability of a availability of a short-term short-term contract link contract link [C, 5-3, 10Mb/s, 30mins, $5] [C-A, 5-3-1, 5Mb/s, 15mins, $1.25+$1.2] 15
Path-V Path-Vector Contract Routing: ector Contract Routing: Micro-level, On-demand, Reactive Micro-level, On-demand, Reactive User initiates… User X wants to know if it can reach 5 with 10-30Mb/s for 15-45mins in a $10 budget [5, 10-30Mb/s, 15-45mins, $10] [5, A, 1-2, 15-30Mb/s, 15-30mins, $8] [5, A-B, 1-2-4, ISP path request t s 15-20Mb/s, 20-30mins, e u q e B r h t a p 2 $4] reply [A-B-C, 1-2-4-5, ISP A 1 4 r e 20Mb/s, 30mins] p User X l y p a t h 3 ISP r e q u e s t C Paths to 5 are found Paths to 5 are found 5 [5, A, 1-3, 5-10Mb/s, and ISP C sends and ISP C sends 15-20mins, $7] replies to the user with replies to the user with two specific contract- two specific contract- [A-C, 1-3-5, 10Mb/s, 15mins] path-vectors. path-vectors. 16
17 Contract Routing over FCT Model Contracting ¡at ¡tier-‑1: ¡long ¡time-‑scale ISP ¡A ¡ ISP ¡B Tier ¡1 ISP ¡C ¡ISP ¡E ¡ISP ¡D Tier ¡2 Tier ¡3 ¡Organization ¡A ¡ Organization ¡B ¡ Organization ¡C ¡ Contract ¡between ¡two ¡tier-‑2 ¡ Contract ¡between ¡two ¡tier-‑3 ¡ networks: ¡medium ¡time-‑scale networks: ¡short ¡time-‑scale
18 Deployment Issues How to motivate ISPs to participate? How to motivate ISPs to participate? ISPs are very protective of their contracting ISPs are very protective of their contracting terms – due to competition. terms – due to competition. But, BGP But, BGP has similar risks too.. has similar risks too.. Observation of opportunity costs Observation of opportunity costs PVCR can be done at will.. PVCR can be done at will.. Not much to loose if ISPs participate with their Not much to loose if ISPs participate with their leftover bandwidth. leftover bandwidth. Monitoring and verification of contracts Monitoring and verification of contracts Who is breaking the e2e performance? Who is breaking the e2e performance? Active measurements can be OK for LSCR, but Active measurements can be OK for LSCR, but PVCR needs lightweight techniques. PVCR needs lightweight techniques.
Recommend
More recommend
