Capacity Planning Understanding capacity planning for system design interviews.
Capacity Planning is the process of determining the resources needed to meet current and future demands. It involves estimating storage, compute, and network requirements.
Avoid over-provisioning : Save costsAvoid under-provisioning : Prevent outagesPlan for growth : Scale proactivelyInterview requirement : Often asked in system design
Metric Question DAU/MAU How many users? QPS How many requests per second? Storage How much data to store? Bandwidth How much data transferred? Memory How much RAM needed?
- 100M daily active users (DAU) - Average user makes 10 requests/day - Average request size: 1 KB - Data retention: 5 years Total requests/day = DAU × requests/user = 100M × 10 = 1 billion/day QPS = 1B / 86,400 seconds ≈ 12,000 QPS (average) Peak QPS = 2-3x average ≈ 30,000 QPS Daily data = requests × size = 1B × 1 KB = 1 TB/day 5-year storage = 1 TB × 365 × 5 = 1.8 PB Incoming = 1B requests × 1 KB = 1 TB/day = ~12 MB/s average Outgoing (if 10 KB response) = 10 TB/day = ~120 MB/s average Metric Value Seconds in a day 86,400 (~100K) Seconds in a month ~2.5M 1 million requests/day ~12 QPS 1 billion requests/day ~12,000 QPS
Resource Typical Value CPU cores 8-64 cores RAM 32-256 GB Storage 1-10 TB SSD Network 1-10 Gbps
Web server: 1,000-10,000 QPS Database: 1,000-5,000 QPS (reads) Cache: 100,000+ QPS (Redis) Required QPS = 30,000 (peak) Server capacity = 5,000 QPS Servers needed = 30,000 / 5,000 = 6 servers Add redundancy (2x) = 12 servers Type Size (typical) User record 1 KB Tweet/Post 250 bytes Image (compressed) 200 KB Video (1 min, compressed) 50 MB Log entry 100 bytes
Actual storage = Raw storage × Replication factor If replication factor = 3: 1 PB raw → 3 PB actual 20% of data serves 80% of requests:
Total data = 1 PB Hot data = 1 PB × 20% = 200 TB If cache hit ratio target = 90%: Cache size ≈ 200 TB Requirements: - 100M URLs created/month - 10:1 read/write ratio - URL stored for 5 years Writes: - 100M / month = 40 URLs/second Reads: - 400 URLs/second Storage: - URL + short code ≈ 500 bytes - 100M × 12 months × 5 years = 6B URLs - 6B × 500 bytes = 3 TB Cache: - 20% hot URLs = 1.2B × 500 bytes = 600 GB
Round numbers for easy math (use powers of 10)
State assumptions clearly
Consider peak vs average traffic
Account for replication and redundancy
Mention growth projections
Know common capacity numbers
Masst Docs