Sifter: An Inversion-Free and Large-Capacity Programmable Packet Scheduler
Talkat 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI '24) · 2024-04-16..04-18·
summary
Sifter claims inversion-free, large-capacity programmable scheduling by combining Sift Sorting coarse buckets (FIFO rotating RCQ) with a small mini PIFO for fine ordering, using a measured speed-up factor (~10 metadata operations per packet transmit) to finish sorting between dequeues. Simulations (NF3) with 256 hosts and >500 flows match ideal PIFO on flow completion time and delay, and an FPGA prototype sustains 100 Gbps line-rate while emitting monotonically increasing ranks (no packet inversions).
AI-written summary · about
chapters (3)
related talks
15mEverything Matters in Programmable Packet Scheduling
13mClubHeap: A High-Speed and Scalable Priority Queue for Programmable Packet Scheduling
22mEiffel: Efficient and Flexible Software Packet Scheduling
23mSP-PIFO: Approximating Push-In First-Out Behaviors using Strict-Priority Queues
16mGearbox: A Hierarchical Packet Scheduler for Approximate Weighted Fair Queuing
22mProgrammable Calendar Queues for High-speed Packet Scheduling
25mApproximating Fair Queueing on Reconfigurable Switches
16mPacket Order Matters! Improving Application Performance by Deliberately Delaying Packets
