Scale Lab

Back-of-the-envelope calculators and reference tables for systems design estimation.

Calculators

Availability Calculator

Convert availability percentage (nines) into allowed downtime per year, month, and day. Includes planned maintenance windows. Essential for SLA discussions in system design.

2 inputs

Bandwidth Calculator

Estimate network bandwidth requirements from QPS and response payload size. Critical for CDN planning, load balancer sizing, and network capacity design.

2 inputs

Caching Impact Calculator

Quantify how caching reduces database load and improves response latency. Demonstrates the power of caching in system design trade-off discussions.

4 inputs

Latency Budget Calculator

Break down an end-to-end latency target into component budgets. Helps identify where time is spent and how much room remains for application logic.

5 inputs

QPS Calculator

Estimate queries per second (QPS) from daily active users and usage patterns. A foundational calculation for capacity planning in system design interviews.

3 inputs

Storage Estimator

Estimate storage requirements over time based on data ingestion rate, record size, and retention policy. Essential for database and object storage capacity planning.

3 inputs

Reference Tables

Latency Numbers Every Programmer Should Know

Key latency figures for reasoning about system performance. Originally compiled by Jeff Dean and updated for modern hardware. Invaluable for back-of-envelope calculations in system design interviews.

Operation	Latency	Nanoseconds(ns)	Category	Notes
L1 cache reference	0.5 ns	0.5	CPU	Fastest memory access; data already in the CPU core's local cache
Branch mispredict	5 ns	5	CPU	Pipeline flush and restart cost when CPU predicts wrong branch
L2 cache reference	7 ns	7	CPU	~14x L1 cache; shared between cores on most architectures
L3 cache reference	20 ns	20	CPU	Shared across all cores; last level before main memory
Mutex lock/unlock	25 ns	25	OS	Uncontended; contended locks can be orders of magnitude slower
Main memory reference (DRAM)	100 ns	100	Memory	~200x L1 cache; DDR5 typical latency
Compress 1 KB with Snappy	3,000 ns (3 us)	3000	CPU	Snappy prioritizes speed over compression ratio
Send 1 KB over 1 Gbps network	10,000 ns (10 us)	10000	Network	Raw bandwidth cost; does not include round-trip overhead
Read 4 KB randomly from SSD	16,000 ns (16 us)	16000	Storage	NVMe SSD; SATA SSD is ~100 us
Read 1 MB sequentially from memory	250,000 ns (250 us)	250000	Memory	~62.5 GB/s effective bandwidth for sequential read
Round trip within same datacenter	500,000 ns (500 us)	500000	Network	~0.5 ms; typical for service-to-service calls within a region
Read 1 MB sequentially from SSD	1,000,000 ns (1 ms)	1000000	Storage	NVMe SSD; ~1 GB/s effective sequential throughput
HDD seek	4,000,000 ns (4 ms)	4000000	Storage	Mechanical disk; seek time dominates random I/O
Read 1 MB sequentially from HDD	5,000,000 ns (5 ms)	5000000	Storage	~200 MB/s sequential; good for throughput, bad for random I/O
Read 1 MB sequentially from network (1 Gbps)	10,000,000 ns (10 ms)	10000000	Network	Dominated by bandwidth limit of 1 Gbps link
Redis GET (in-memory, local datacenter)	500,000 ns (0.5 ms)	500000	Data Store	Typical for a cache hit; includes network round trip
Simple database query (indexed)	5,000,000 ns (5 ms)	5000000	Data Store	Simple key lookup on indexed column; varies widely by engine
Send packet CA -> Netherlands -> CA	150,000,000 ns (150 ms)	150000000	Network	Transatlantic round trip; limited by speed of light in fiber
TLS handshake	250,000,000 ns (250 ms)	250000000	Network	TLS 1.2 full handshake (2 round trips); TLS 1.3 is ~1 RTT
Retry with exponential backoff (1st retry)	1,000,000,000 ns (1 s)	1000000000	Application	Common starting point for retry delays; doubles each attempt

Powers of 2

Essential reference for back-of-envelope calculations. Every systems designer should know these by heart: how many bytes in a kilobyte, megabyte, gigabyte, and beyond.

Power	Exact Value	Approx Value	Common Name
2^0	1	1
2^1	2	2
2^2	4	4
2^3	8	8
2^4	16	16
2^5	32	32
2^6	64	64
2^7	128	128
2^8	256	256
2^9	512	512
2^10	1,024	~1 Thousand	1 KB (Kilobyte)
2^11	2,048	~2 Thousand	2 KB
2^12	4,096	~4 Thousand	4 KB
2^13	8,192	~8 Thousand	8 KB
2^14	16,384	~16 Thousand	16 KB
2^15	32,768	~32 Thousand	32 KB
2^16	65,536	~64 Thousand	64 KB
2^17	131,072	~128 Thousand	128 KB
2^18	262,144	~256 Thousand	256 KB
2^19	524,288	~512 Thousand	512 KB
2^20	1,048,576	~1 Million	1 MB (Megabyte)
2^21	2,097,152	~2 Million	2 MB
2^22	4,194,304	~4 Million	4 MB
2^23	8,388,608	~8 Million	8 MB
2^24	16,777,216	~16 Million	16 MB
2^25	33,554,432	~32 Million	32 MB
2^26	67,108,864	~64 Million	64 MB
2^27	134,217,728	~128 Million	128 MB
2^28	268,435,456	~256 Million	256 MB
2^29	536,870,912	~512 Million	512 MB
2^30	1,073,741,824	~1 Billion	1 GB (Gigabyte)
2^31	2,147,483,648	~2 Billion	2 GB (max signed 32-bit int)
2^32	4,294,967,296	~4 Billion	4 GB (max unsigned 32-bit int)
2^33	8,589,934,592	~8 Billion	8 GB
2^34	17,179,869,184	~16 Billion	16 GB
2^35	34,359,738,368	~32 Billion	32 GB
2^36	68,719,476,736	~64 Billion	64 GB
2^37	137,438,953,472	~128 Billion	128 GB
2^38	274,877,906,944	~256 Billion	256 GB
2^39	549,755,813,888	~512 Billion	512 GB
2^40	1,099,511,627,776	~1 Trillion	1 TB (Terabyte)