Download presentation
Presentation is loading. Please wait.
Published byMia miah Crozier Modified over 9 years ago
1
Clean Slate Design for the Internet http://cleanslate.stanford.edu Designing a Predictable Backbone Network with Valiant Load Balancing NSF 100 x 100 Clean Slate Program http://100x100network.org Nick McKeown Stanford University All the hard stuff was done by: Rui Zhang-Shen
2
Backbone network design using VLB Backbone network 4 5 1 2 3
3
Backbone network design using VLB US Backbone Networks: Observations ~50 nodes interconnected by long-haul optical links Increasingly rich mesh topology –Built over mesh of WDM or TDM circuits and switches –Reduce hop count and delay –Fault tolerance –Load balancing Low utilization—links over-provisioned –Uncertainty in traffic matrix the network is designed for –Headroom for future growth –Prepare to take over when links or routers fail –Minimize congestion and delay variation
4
Backbone network design using VLB Traffic Matrices 4 5 1 2 3 From To Traffic matrix is hard to predict Regional Node i riri Back- bone r i is easier to predict and has to be predicted anyway riri
5
Backbone network design using VLB What fraction of traffic matrices can they support? Abilene Verio AT&T Sprint 80% utilization: 0.026% 67% utilization: 0.66% 80% utilization: 0.0003% 67% utilization: 1.1% 80% utilization: 0.00008% 67% utilization: 0.09% 80% utilization: 0.0009% 67% utilization: 0.026% Verio, AT&T, and Sprint topologies courtesy of RocketFuel
6
Backbone network design using VLB Desired Characteristics Dependable –Continues to operate when traffic patterns change in the short and long term –Continues to operate under failure –Recovers quickly Efficient –And at no extra cost
7
Backbone network design using VLB Why is this hard? 1 2 3 N … 4 r r r r r r r r r r r
8
Backbone network design using VLB Why is this hard? 1 2 3 N … 4 r r r r r r r r r r r Nr
9
Backbone network design using VLB Our Approach The operator already estimates r i –Requires only local knowledge of users and market estimates Use Valiant Load Balancing (VLB) –Supports all traffic matrices History –L. G. Valiant, G. Brebner, 1981-82 Parallel communication Statistical delay guarantee –C.-S. Chang, etc.; I. Keslassy etc., 2001-05 Switch scheduling Throughput guarantee Optimality
10
Backbone network design using VLB Valiant Load-Balancing 1 2 3 N … 4 r r r r r r
11
Backbone network design using VLB Valiant Load-Balancing 1 2 3 N … 4 r r r r 2r/N r r 24 In practice : The mesh could be a mesh of lambdas or TDM circuits Send on direct path, and only spread when network is congested.
12
Backbone network design using VLB Aside: Routers based on VLB Can you build a router switched backplane based on VLB? Appealing possibilities 100% throughput for any arrival pattern No per-packet arbitration and scheduling Passive switch fabric consumes almost zero power Linecards 5556 12 40 x 40 MEMS Switch Rack < 100W “Scaling Routers using Optics” Sigcomm 2003
13
Backbone network design using VLB Failures Node failures –Takes away corresponding links and traffic –Still a full mesh network Links failures –Asymmetric network –Many scenarios 1 2 3 N … 4
14
Backbone network design using VLB Fault Tolerance Load balance traffic over available paths To tolerate any k link or router failures, sufficient to increase the capacity each link by Example: A 50 node network requires 11% more capacity to withstand any 5 failures.
15
Backbone network design using VLB Heterogeneous Network 1 2 3 N … 4 r1r1 r4r4 r3r3 riri rNrN r2r2 R = i r i Homogeneous: c = 2r/N c ij r i r j c ij = 2r i r j /R Gravity Configuration
16
Backbone network design using VLB Heterogeneous Network As before, the total capacity we need with VLB is twice what we’d need if we knew the traffic matrix (and it was static). With oblivious routing we need an extra capacity.
17
Backbone network design using VLB Is VLB efficient? Not knowing the traffic matrix means we need a total capacity 2-times larger than if we did. But we never know the traffic matrix, and it changes. So the cost is surprisingly small. Anecdotally, a network that can support all traffic matrices and behaves predictably on failure requires less capacity than existing networks.
18
Backbone network design using VLB Interconnecting Backbones Peering parameters: –R p is maximum peering traffic –q i ¸0 for peering nodes, q i =0 for non-peering nodes, i q i =1 Peering link capacity: R p q i 1 2 3 N … 4 2 1 N 3 4 … RpRp q2q2 q3q3 q4q4
19
Backbone network design using VLB Within a VLB Network Assume peering condition is fixed –Given: R p ; q i –Variables: p i Spread traffic over the peering links
20
Backbone network design using VLB Spread over peering links c ij = r i p j + r j p i + min(r i,R p )(max(p j,q j )-p j ) + min(r j,R p )(max(p i,q i )-p i ) If R p > r i, optimal solution: p i = q i ; c ij = r i q j + r j q i Efficient use of peering links Supports all traffic matrices as before 1 2 3 N … 4 2 1 N 3 4 … 111
21
Backbone network design using VLB Other questions Delay-sensitive applications –How much does it matter? –It may matter for interactive voice, video, gaming –Dealing with it: “Express paths”,Adaptive load-balancing
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.