SystemDesignDraw Logo
SystemDesignDraw

Pizarra de Arquitectura y Diagramas

Referencia Técnica de Ingeniería de Sistemas

Guía Rápida de Arquitectura de Sistemas & Calculadora

Fórmulas de capacidad, cálculos de ancho de banda y latencias de Peter Norvig, disponibilidad de SLAs de 9s y teoremas distribuidos de Martin Kleppmann y Jeff Dean.

Calculadoras Interactivas de Arquitectura

Modela el tráfico, la memoria y la disponibilidad de sistemas en producción:

e.g. 10 Million users
Clicks, API calls, reads
Incoming payload size
Outgoing JSON/HTML
Workload split
Average QPS3,472Reqs / second
QPS Pico (2.5x)8,680Surge headroom
QPS Lectura vs Escritura3.1K R / 347 WStorage writes only
Ancho de Banda Saliente416.6 MbpsOutbound wire rate
Interfaz de Red (NIC)1 Gbps NIC70% safe limit
Storage (5 Years)109.5 TBWrites compounded
Hardware Scale Benchmarks

Latency Numbers Every Engineer Must Know

Understanding relative orders of magnitude between CPU, RAM, NVMe, and network hops:

👈 Swipe horizontally to view full table 👉
Hardware OperationDean 2009 BaselineModern 2026 DatacenterHuman Scale (1ns = 1s)Architectural Rule of Thumb
L1 CPU Cache Reference0.5 ns0.7–1.0 ns1 secondSub-nanosecond compute; use cache-friendly contiguous arrays.
Branch Mispredict5 ns2–3 ns3 secondsFlushes CPU pipeline; write branchless code in hot paths.
Main RAM Reference (DDR4/5)100 ns50–80 ns1.2 minutes100x slower than L1; pointer-heavy trees cause cache-miss stalls.
Snappy / ZSTD Compression (1 KB)10 µs1.5–3.0 µs50 minutesCPU compression is 10x faster than transmitting uncompressed wire bytes.
Send 2 KB over 1 Gbps Network20 µs20 µs5.5 hoursWire serialization delay; 25/100G NICs reduce this to ~200 ns.
NVMe SSD Random 4 KB Read~100 µs10–25 µs7 hoursFastest disk storage; ~1,000x faster than rotational HDD seek.
Same-Datacenter RTT (Same AZ)500 µs (0.5 ms)150–300 µs3.5 days3,000x slower than local RAM. Batch remote calls into single MGETs.
Rotational HDD Seek10 ms8–10 ms4 monthsMechanical head motion caps throughput at ~120 IOPS per disk.
Cross-Continental RTT (SF to NYC)~40 ms40 ms1.3 yearsSpeed of light in fiber (200,000 km/s). Requires edge CDNs.
Transpacific RTT (CA to Europe/Asia)150 ms150 ms4.75 yearsMulti-region synchronous database writes cannot bypass this latency.
Abadi Theorem (Beyond CAP)

PACELC Database Decision Matrix

CAP only applies during partitions. PACELC explains behavior in normal operations (Latency vs Consistency):

PC / ECPartition: Consistency • Normal: Consistency

Strict Consensus Databases

Google Cloud Spanner, CockroachDB, Apache HBase, Bigtable

Rejects writes on minority partitions; waits for consensus replication before returning.Best for: Financial ledgers, stock trading, inventory reservation, strict ACID compliance.

PC / ELPartition: Consistency • Normal: Low Latency

Primary-Follower Stores

MongoDB (w:1), PostgreSQL / MySQL (Async Replication)

Preserves linearizability during network splits, but optimizes for sub-millisecond local writes during normal health.Best for: User profiles, e-commerce catalogs, SaaS dashboards.

PA / ELPartition: Availability • Normal: Low Latency

Eventual-Consistency Leaderless

Apache Cassandra, Amazon DynamoDB, ScyllaDB, Riak

Accepts writes on all partitions; replicates asynchronously in background. Ultra-fast write latency.Best for: High-volume IoT telemetry, shopping cart items, chat history, metrics logs.

PA / ECPartition: Availability • Normal: Consistency

Monotonic Timeline Stores

Yahoo! PNUTS (Sherpa), Hazelcast Split-Brain Modes

Guarantees per-record monotonic timeline order under normal conditions while remaining available during datacenter partitions.Best for: Social timeline feeds with master record leases.

RFC 9110 & Distributed APIs

HTTP Idempotency & Distributed Status Codes

How distributed gateways and microservices handle retries, timeouts, and concurrency conflicts:

Safe vs Idempotent HTTP Methods

👈 Swipe 👉
MethodSafe?Idempotent?Auto-Retry Rule
GET / HEADYesYesSafe to replay anytime.
PUTNoYesReplaces full state; safe to retry.
DELETENoYesResource stays deleted; safe to retry.
POST / PATCHNoNONEVER retry without Idempotency-Key!

!Distributed System Error Triage

409 Conflict: OCC version / ETag mismatch or duplicate in-flight idempotency key. Do NOT retry blindly; re-read latest state and diff.
429 Too Many Requests: Token bucket saturated. Inspect Retry-After: N header; backoff with full jitter.
504 Gateway Timeout: Indeterminate State. Downstream database write may have committed before timeout. Never blindly replay payment POSTs!

Back-of-the-Envelope Mental Math Shortcuts

Alex Xu & Google SRE rules to calculate interview numbers in seconds without a calculator:

Time Constants

• 1 Day = 86,400 s ≈ 100,000 s (10⁵ seconds)

• 1 Year = 31.5 × 10⁶ seconds

• 1 Million req/day ≈ 12 QPS

• 100 Million req/day ≈ 1,200 QPS

Data Size Equivalents

• 2¹⁰ ≈ 1,000 = 1 KB (Kilobyte)

• 2²⁰ ≈ 1,000,000 = 1 MB (Megabyte)

• 2³⁰ ≈ 10⁹ = 1 GB (Gigabyte)

• 2⁴⁰ ≈ 10¹² = 1 TB (Terabyte)

5-Year Compounding Shortcut

• 365 × 5 = 1,825 ≈ 2,000×

• Daily Storage × 2,000 ≈ 5-Year Storage

*The extra 175× naturally accommodates index B-tree overhead, write replication, and organic user growth!

Ready to diagram your production system?

Apply these calculations to our interactive Excalidraw canvas or explore real-world blueprints.