Scheduler

Scheduler selects the optimal parent peer for current peer to be downloaded and triggers the seed peer back-to-source download or Client back-to-source download at the appropriate time.

Features

• Based on the multi-feature intelligent scheduling system selects the optimal parent peer.
• Build a scheduling directed acyclic graph for the P2P cluster.
• Remove abnormal peer based on peer multi-feature evaluation results.
• In the case of scheduling failure, notice peer back-to-source download.
• Provide metadata storage to support file writing and seeding.

Scheduling

Scheduler maintains task, peer and host resources.

• Peer: a download task for Client
• Host: host information for Client, host and peer have a 1:N relationship
• Task: a download task, task and peer have a 1:N relationship

The scheduling process is actually to build a directed acyclic graph according to the peer's load.

Load-Aware Scheduling Algorithm

A two-stage scheduling algorithm combining central scheduling with node-level secondary scheduling to optimize P2P download performance based on real-time load awareness.

Architecture

Stage 1: Central Scheduling

The Scheduler evaluates and returns the top N parent for concurrent piece downloads.

Total Score Formula

TotalScore = (IDCAffinityScore × 0.2) + (LocationAffinityScore × 0.1) + (LoadQualityScore × 0.6) + (HostTypeScore × 0.1)

Name	Weight	Description
IDC Affinity Score	0.2	Data center proximity preference
Location Affinity Score	0.1	Geographic location matching
Load Quality Score	0.6	Load evaluation
Host Type Score	0.1	Node type preference (seed, peer, etc.)

Load Quality Score

LoadQualityScore = (PeakBandwidthUsage × 0.5) + (BandwidthDurationRatio × 0.3) + (ConcurrentEfficiency × 0.2)

Name	Weight	Description
Peak Bandwidth Usage	0.5
Bandwidth Duration Ratio	0.3
Concurrent Efficiency	0.2

Stage 2: Secondary Scheduling

Real-time parent selection performed by the peer for each piece based on load quality.

While the central scheduler provides initial candidate selection, the Parent Selector enables dynamic adaptation to:

• Real-time load: Network conditions fluctuate rapidly during downloads.
• Reduced central load: Distributes scheduling decisions across peers.
• Lower latency: Local decisions avoid round-trip delays to central scheduler.