View Count on a Video/Post

Dedup, bot-filter, batched aggregation.

SavedSaved on this device

The problem

Dedup, bot-filter, batched aggregation.

Count the views on a video or post. It sounds like count++; it is one of the harder write-heavy distributed-systems problems in practice. The work is dominated by three pressures the naive version ignores: dedup (the same viewer reloading must not inflate the number), bot / invalid-traffic filtering (a large fraction of raw "views" are not humans, and the public number must be reported net of them), and batched aggregation (you cannot synchronously increment a row per view at 250k views/sec — you log an idempotent event and roll it up asynchronously). The system is also read-dominated for the display of the count and hot-keyed on the write of a viral video, so it pulls in opposite directions at once.

This canonical models the main flows: the ingest/write path, the read path that displays the count, the speed-layer aggregation that produces the fast approximate number, and the batch "verified-views" layer that produces the authoritative one — the layer behind YouTube's famous 301+ freeze and the reason counts sometimes go down.

Reading: Netflix Distributed Counter Abstraction (TechBlog, 2024) · Cloudflare HTTP analytics @6M req/s on ClickHouse · Reddit View Counting (HLL + Kafka + Cassandra) · Twitter TSAR / Summingbird + Manhattan · LinkedIn 'Who Viewed Your Profile' on Apache Pinot · Ably: Cassandra counter columns — nice in theory, hazardous in practice · Google Ad Manager GIVT/SIVT invalid-traffic methodology

fire-and-forget event log
idempotent counter rollup
HyperLogLog uniques
bot / invalid-traffic filtering
sharded counters & hot keys
lambda batch reconciliation
monotonic serving

01Clarifications

What would you ask before drawing a single box?

Ambiguity you would resolve with the interviewer: scope, scale, who uses it, what counts as done.

About View Count on a Video/Post

Dedup, bot-filter, batched aggregation.

Difficulty: intermediate
Time: about 45 minutes
Stages: 10
Topic: Feeds, Timelines, Counters & Ranking

How this problem is worked

Ten stages, from the questions you would ask an interviewer to the trade-offs you would defend. Each asks one question, and the simulator runs the architecture you draw against the requirements you wrote.

01ClarificationsWhat would you ask before drawing a single box?
02Functional reqsWhat must this system actually do?
03Non-functionalWhat must it promise about speed, uptime and correctness?
04Capacity estimationHow much load and data does this have to hold?
05API designWhat does the outside world call, and what comes back?
06Data modelWhat gets stored, and what is it looked up by?
07Use-case breakdownHow does each requirement actually get served?
08High-level designWhich components handle a request, and in what order?
09Deep divesWhich part breaks first, and what do you do about it?
10Trade-offsWhat did this design cost, and what breaks at 10×?

Primary sources for this problem

Netflix Distributed Counter Abstraction (TechBlog, 2024)
Cloudflare HTTP analytics @6M req/s on ClickHouse
Reddit View Counting (HLL + Kafka + Cassandra)
Twitter TSAR / Summingbird + Manhattan
LinkedIn 'Who Viewed Your Profile' on Apache Pinot
Ably: Cassandra counter columns — nice in theory, hazardous in practice
Google Ad Manager GIVT/SIVT invalid-traffic methodology

More in Feeds, Timelines, Counters & Ranking

What to show and in what order: fanout on write versus read, hot/top/new scoring, approximate counters, trending, and recommendation.

Browse the full problem catalog, or see what the simulator does and does not model.