Chain Drive Architecture Demo
300 equal website links — idle supports busy — priority-stacked work — self-repair around failed links — OCC symbolic packets cut repeat transfer.
1. Chain Drive Architecture
In the Chain Drive model, every website in the ecosystem is an equal link in a continuous circular chain. There is no master and no slaves. Work flows around the ring in priority order. At any moment, a busy or overloaded site can pull capacity from idle neighbors; a failed link is absorbed by its neighbors so the chain keeps running. Repeated structural payloads ship as small OneCharacterCode symbolic packets instead of full re-downloads.
2. 300 Equal Website Links
The simulation builds 300 synthetic ecosystem websites — each with a SiteID, domain, priority level, current load, queue type, dependencies, shared customers, status, and counters for work completed / support given / support received. The full initial configuration is downloadable below as chain-drive-simulation.json.
3. Priority Stack Order
Work in the chain is dispatched in this order. Higher-priority items always run before lower-priority items on the same link. Sites with nothing in their top queue fall down the stack until they find work; sites with no work at all rest and become available to support overloaded neighbors.
- Accounts — identity / billing / membership writes
- Security alerts — abuse, fraud, intrusion signals
- Payments / sales — order, checkout, refund flows
- Customer updates — profile, preference, opt-in changes
- Inventory updates — price / stock / SKU propagation
- Website content updates — pages, blogs, landing copy
- SEO indexing — sitemap regeneration, schema, internal links
- Backup tasks — rolling site / DB snapshots
- Archive tasks — long-term cold storage
- Rest / idle — available capacity, supports busier links
4. Load Balancing Demo
In Scenario B, site #47 receives an event spike that adds ~80% load on top of its existing capacity. The visualization shows it flashing red (overloaded), and nearby idle / archive / SEO sites flash blue as they absorb the excess work. The counter row above the chain shows support transfers performed climbing in real time, and the overloaded after balancing counter drops back to 0 once the chain settles.
5. Self-Repair Demo
In Scenario D, three important sites (#133, #220, #270) fail simultaneously. Each failed node is shown in red. Its two ring-neighbors flash purple as they absorb its scheduled work for the rest of the cycle. The failed links recovered counter tracks how many absorption events occurred. The chain continues running without pausing — this is the key difference from a star-topology architecture where a single failure can stop the system.
6. OCC Symbolic Packet Savings
Every cross-link transfer in the simulation reports two byte counts: normal transfer bytes (what a full payload push would cost) and symbolic packet bytes (what an OCC-style symbolic reference would cost, using a shared installed dictionary). Because UI chrome, JSON schema keys, log frames, and product templates are heavily repeated across the ecosystem, the symbolic packet size is a small fraction of the full transfer size.
Aggregate row will appear once results load.
7. Simulation Results
| Scenario | Tasks | Avg load % | Overld before | Overld after | Idle reused | Supports | Fail recov. | Normal bytes | Symbolic bytes | % saved |
|---|---|---|---|---|---|---|---|---|---|---|
| Loading simulation results… | ||||||||||
8. Honest Limitations
- This is a simulation, not a live production deployment. No real cross-site network traffic happens here.
- The 300 nodes, their domains, priority assignments, and dependencies are deterministic synthetic data generated from a fixed seed. They do not represent any specific live website.
- Symbolic-packet vs normal-transfer byte sizes are modeled from priority-class defaults; they are not measured packet captures. In production these numbers would be measured against the OCC engine on the wire.
- The animated SVG is a cosmetic visualization of the simulation state. It is synchronized to the scenario but its per-frame colors are approximated; the authoritative numbers live in
chain-drive-results.json. - Self-repair here absorbs work into neighbor capacity within one cycle; a real deployment would also queue retries, handle network partitions, and run health probes.
- Priority-stack ordering is implemented as a simple greedy queue. Real production would add fairness, starvation prevention, and tenant quotas.
- OCC dictionary install cost is not modeled here in the wire-bytes figure (it appears in the V4 benchmark). The Chain Drive demo focuses on the runtime savings after install.