Using Redis to Count Website Visits: Hash, Bitset, and Probabilistic Algorithms

During a Pinduoduo backend interview, candidates are asked how to use Redis to count the number of visits to a specific webpage for a site with hundreds of millions of users.

Using Hash

Redis hashes store key‑field‑value pairs; each user visit can be recorded with HSET where the Redis key combines the page URI and date, the field is the user ID (or a generated identifier for anonymous users), and the value is set to 1. The total daily visits are obtained with HLEN . Advantages: simple, easy to implement, accurate results. Disadvantages: high memory consumption and performance degradation as the number of keys grows, making it unsuitable for massive traffic.

Using Bitset

A bitset represents each user as a single bit, drastically reducing memory usage (e.g., 100 million users require only about 12 MB). Redis provides SETBIT to mark a user’s visit, GETBIT to query a specific user, and BITCOUNT to count the total set bits for the day. Advantages: very low memory usage and fast queries. Disadvantages: potential memory waste for sparse data and the need for an extra mapping for anonymous users.

Using Probabilistic Algorithm

For approximate counts, Redis’s HyperLogLog can be used. Each visit triggers PFADD to add the user identifier to the HLL structure, and the daily unique visitor count is retrieved with PFCOUNT . This method uses roughly 12 KB per key, making it ideal for extremely large traffic, though it introduces about a 0.81 % error margin. Advantages: minimal memory footprint, suitable for billions of users. Disadvantages: cannot query individual users and results have a small statistical error.

Source: https://url.cn/5tQPEQg