今日固定时间表
三块窗口固定：08:50-10:55 系统设计，14:30-16:15 算法，20:30-21:15 口述。
| 08:50-09:15 | 08:50-10:55 system design Bad/Good/Great - Bad: query max then write; Good: DB cached max; Great: per-auction log + conditional highest + durable queue。 |
|---|---|
| 09:15-09:50 | 08:50-10:55 system design 精确原文 - 读 Online Auction Potential Deep Dives、Some additional deep dives 和 level expectations。 |
| 09:50-10:25 | 08:50-10:55 system design 故障时序 - bid accepted ACK 丢、consumer crash、SSE fanout lag、dynamic close 与 late bid race。 |
| 10:25-10:55 | 08:50-10:55 system design 恢复和指标 - Kafka lag, accepted-bid duplicate rate, highest projection staleness, close mismatch sample, fanout p99。 |
| 14:30-16:15 | NeetCode Tag · Binary Search 三道题，每题 30m solve + 5m evidence，不能改 tag。 |
| 20:30-21:15 | 20:30-21:15 Frontier radar deliberate omission + Auction follow-up recording。 |
算法模块
连续 NeetCode Binary Search；每天 exactly 3 slots。
| # | Problem / mode | Pattern | Invariant | Bug risk | Time | Space |
|---|---|---|---|---|---|---|
| 1 | Find the Index of the Large Integer NEW · 30m solve + 5m evidence | interactive compare halves | The heavy index remains inside the selected half. | Uneven split loses the middle element. | O(log n) | O(1) |
| 2 | Divide Chocolate NEW · 30m solve + 5m evidence | maximize minimum feasible value | If min sweetness x is feasible, any smaller x is feasible. | Forgetting you need k+1 pieces; greedy split condition. | O(n log sum) | O(1) |
| 3 | Maximum Average Subarray II NEW · 30m solve + 5m evidence | binary answer with prefix feasibility | avg x is feasible if adjusted subarray length >= k has nonnegative sum. | Precision tolerance and min-prefix index off-by-one. | O(n log R) | O(n) or O(1) |
Hello Interview 精确目录
保留 live canonical heading title；day page 使用 direct section anchors。
| 项目 | 精确 canonical section | Disposition | 今天怎么用 |
|---|---|---|---|
| Online Auction | Potential Deep Dives | deep dive | contention, durability, real-time updates and scale |
| Online Auction | 1) How can we ensure strong consistency for bids? | deep dive | avoid stale max-bid acceptance and define winner order |
| Online Auction | 2) How can we ensure that the system is fault tolerant and durable? | deep dive | durable queue and no dropped bids |
| Online Auction | 3) How can we ensure that the system displays the current highest bid in real-time? | deep dive | SSE/current highest bid propagation |
| Online Auction | 4) How can we ensure that the system scales to support 10M concurrent auctions? | deep dive | partition by auctionId, shard writes and fanout |
| Online Auction | Some additional deep dives you might consider | deep dive | dynamic end times, purchasing, live bid history |
| Online Auction | What is Expected at Each Level? | level expectation | calibrate full-coverage follow-up depth |
| Online Auction | Mid-level | level expectation | basic HLD and consistency response |
| Online Auction | Senior | level expectation | lead consistency and real-time bid discussion |
| Online Auction | Staff | level expectation | proactive closing, fraud, failure and durable fairness |
DDIA 精确决策卡
只使用 Ch8 和 Ch10 中直接解决今天设计决定的小节。
| 项目 | 项目问题 | 精确 DDIA subsection | 改变的系统决定 |
|---|---|---|---|
| Online Auction | How do equal/near-close bids get a deterministic order? | Ch10 - 全序广播 | Per-auction log order defines tie-break and close eligibility; notification can lag but winner computation uses the ordered fact log. |
| Online Auction | How do we compute final winner without phantom late bids? | Ch8 - 可串行化快照隔离（SSI） | Close transaction checks the auction end condition and accepted bid range under serializable-equivalent isolation. |
三域 trace 任务
Winning transaction标出唯一线性化点：seat hold/sold、accepted bid at close、brokerage order/execution state。
Duplicate/retry failure给出 stable idempotency key、重复返回语义、旧 owner/stale version 被拒绝的证据。
Projection boundary说明哪些读可以陈旧：seat map、highest bid display、market price UI；哪些写不能陈旧。
本日 artifacts
项目脚本、音频、PDF 和 mock evidence 从当天直接打开。
今日产出、修复与英语 recall
| 课程角色 | FULL · deep-dive follow-up + frontier omission |
|---|---|
| 必须交付 | Online Auction follow-up sheet：strong consistency, durability, real-time highest bid, 10M scale, dynamic close。 |
| 修复规则 | 如果把实时显示和 winner computation 混为一谈，分开 ordered ledger 与 SSE projection。 |
| English recall | English recall: Notification lag is tolerable; accepting the wrong winning bid is not. |
Detailed lecture notes, audio, recall scripts, PDFs, Staff Q&A, and mock packs are archived locally and are intentionally not published on this site.