A Fast Solver-Free Algorithm for Traffic Engineering in Large-Scale Data Center Network
summary
Sequential Source-Destination Optimization (SSDO) decomposes DCN TE into per source‑destination subproblems and replaces per‑subproblem LP solves with a solver-free binary search (Balanced Binary Search Method) that finds split ratios minimizing MLU; it orders updates by the current hottest link to converge fast. On Meta topologies it reports 65% MLU reduction vs TEAL and 60% vs POP while delivering a 12× speedup over POP, handles ~90,000‑variable scales, supports hot/cold/hybrid seeds, and yields interruptible, near‑LP solutions.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:44Optimization challenges
- 3:53Sequential vs parallel methods
- 5:17SSDO algorithm core
- 7:49Balancing split ratios
- 9:37SSDO deployment
- 10:40SSDO solution quality
- 12:30Conclusion
related talks
17mDOTE: Rethinking (Predictive) WAN Traffic Engineering
26mParallel Declarative Solutions of Sequencing Problems using Multi-valued Decisio
53mOptimal Power Flow: Relaxation, Online Algorithm, Fast Dynamics
16mDecouple and Decompose: Scaling Resource Allocation with DeDe
14mEfficient Direct-Connect Topologies for Collective Communications
19mSol: Fast Distributed Computation Over Slow Networks
17mSolving Max-Min Fair Resource Allocations Quickly on Large Graphs
19mLook Before, Before You Leap: Online Vector Load Balancing with Few Reassignments