Metrics Collection — System Design Academy

Complete System Design Tutorial

The design separates the latency-sensitive online path, the correctness-critical durable transition, and asynchronous projection or repair work. Each chapter below explains one decision shown in the diagrams.

1. Requirements and scope

Functional requirements

Execute the primary metrics collection workflow and return an observable result.
Expose the concrete operations shown below with stable retry identities.
Allow operators to inspect, disable, repair, or replay ambiguous state.
Keep derived analytics and notifications off the correctness-critical path.

Non-functional requirements

Design for 100M samples/s · 15 s scrape · trillions of points.
Protect the named business invariant under concurrency and partial failure.
Scale by an explicit aggregate/partition key without one global mutex.
Degrade intentionally when cache, queue, store, or regional ownership is unavailable.

Primary invariant: each accepted metrics collection operation has one stable result and retries do not repeat its business effect.

↑ top

2. Capacity assumptions

State assumptions before choosing technology. The supplied workload is 100M samples/s · 15 s scrape · trillions of points. Convert it into peak throughput, retained records, bandwidth, and failure budget; then connect each number to a component.

Dimension	Working estimate	Design implication
Traffic	100M samples/s · 15 s scrape · trillions of points	Separate online compute, durable state, and asynchronous consumers.
Peak factor	5–10× average	Admission control, autoscaling headroom, and hot-key protection.
Stored state	write rate × retention × record bytes	Partition count, indexes, compaction, backup, and archival tier.
Network	peak operations × payload size	Batching, compression, edge placement, and egress budget.
Availability	99.9–99.99%	Multi-zone replicas, tested failover, and dependency error budgets.
Recovery	RPO/RTO by invariant	Ack boundary, replay retention, reconciliation, and operator runbooks.

Do not cargo-cult the numbers: show the arithmetic, name the uncertain variable, and say which production metric will replace the interview estimate.

↑ top

3. API and data model

POST /api/v1/writecompressed sample batch

GET /api/v1/query_rangebounded time-series query

POST /v1/rulesrecording/alert rule

GET /v1/cardinalitybudget diagnostics

Authoritative records and indexes

series(fingerprint PK, metric, sorted_labels, tenant)

sample(fingerprint, timestamp, value)

block(tenant, min_time, max_time, level, object_uri)

rule(rule_id, expression, interval, for_duration, version)

Command and event envelopes

# Command envelope
operation: POST /api/v1/write
request_id: req_01J...
idempotency_key: idem_7f2...
aggregate_key: metrics-system_123
expected_version: 17
payload: {...}

# Durable event
event_id: evt_01J...
aggregate_key: metrics-system_123
event_type: metrics_system_committed
schema_version: 3
trace_id: tr_8ac...

The authoritative aggregate and idempotency record share a transactional or conditional-write boundary. Event streams, search indexes, caches, analytics, and dashboards are rebuildable projections.

↑ top

4. Why this architecture and what it costs

The boxes are not a shopping list. Each boundary exists because it owns a different invariant, scaling dimension, data lifecycle, security policy, or failure domain. The table makes those choices explicit and shows when the design should remain simpler.

Component	Owned responsibility	Why separate it	Rejected alternative	Cost and evolution trigger
Instrumented Service	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Discover targets and scrape with bounded concurrency or accept authenticated remote write.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents the external caller from also owning the contract and failure behavior of Scraper / Agent.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Unbounded label creates millions of series” to the external caller.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Instrumented Service boundary before extracting or scaling it.
Scraper / Agent	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Normalize labels, reject cardinality violations, and append batches to a durable log/WAL.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Instrumented Service from also owning the contract and failure behavior of Service Discovery.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Late sample targets compacted block” to Instrumented Service.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Scraper / Agent boundary before extracting or scaling it.
Service Discovery	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Route each fingerprint to one ingest owner and compact head chunks into immutable blocks.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Scraper / Agent from also owning the contract and failure behavior of Remote Write Gateway.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Query scans too many series” to Scraper / Agent.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Service Discovery boundary before extracting or scaling it.
Remote Write Gateway	Owns authentication, validation, quotas, deadlines, and routing. In this design it participates in: Query frontend splits time/shards, deduplicates replicas, and caches partial results.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Service Discovery from also owning the contract and failure behavior of Label Guard.	Put protocol, authentication, quotas, and business transitions in one process. Acceptable at low scale, but it couples external traffic risk to domain releases. Here that simpler choice would couple “Replica dedupe hides missing data” to Service Discovery.	Costs another hop, policy distribution, and a high-blast-radius fleet. Introduce it when many clients or services need one protocol and protection boundary. Measure the latency, saturation, lag, or mismatch at the Remote Write Gateway boundary before extracting or scaling it.
Label Guard	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Rule evaluators checkpoint durable timestamps and send deduplicated alert state.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Remote Write Gateway from also owning the contract and failure behavior of Durable Sample Log.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Rule evaluation misses interval during failover” to Remote Write Gateway.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Label Guard boundary before extracting or scaling it.
Durable Sample Log	Preserves durable ordering, replay, and consumer checkpoints. In this design it participates in: Discover targets and scrape with bounded concurrency or accept authenticated remote write.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Label Guard from also owning the contract and failure behavior of Series Router.	Call every downstream consumer synchronously. That couples user latency and availability to analytics, notification, indexing, and audit dependencies. Here that simpler choice would couple “Unbounded label creates millions of series” to Label Guard.	Costs eventual consistency, schema evolution, replay tooling, idempotent consumers, and lag operations. Earn it when bursts or downstream failure must not block acknowledgment. Measure the latency, saturation, lag, or mismatch at the Durable Sample Log boundary before extracting or scaling it.
Series Router	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Normalize labels, reject cardinality violations, and append batches to a durable log/WAL.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Durable Sample Log from also owning the contract and failure behavior of TSDB Ingest Node.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Late sample targets compacted block” to Durable Sample Log.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Series Router boundary before extracting or scaling it.
TSDB Ingest Node	Owns authoritative versioned state or immutable evidence. In this design it participates in: Route each fingerprint to one ingest owner and compact head chunks into immutable blocks.	Defines the source of truth and commit boundary. It prevents Series Router from also owning the contract and failure behavior of WAL / Head Block.	Let each service persist its own interpretation. That removes one clear commit boundary and makes retries, support, and reconciliation ambiguous. Here that simpler choice would couple “Query scans too many series” to Series Router.	Costs schema migration, replication, backup, and shard ownership. Split storage only when access patterns, retention, or compliance boundaries materially differ. Measure the latency, saturation, lag, or mismatch at the TSDB Ingest Node boundary before extracting or scaling it.
WAL / Head Block	Owns authoritative versioned state or immutable evidence. In this design it participates in: Query frontend splits time/shards, deduplicates replicas, and caches partial results.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents TSDB Ingest Node from also owning the contract and failure behavior of Object Block Store.	Let each service persist its own interpretation. That removes one clear commit boundary and makes retries, support, and reconciliation ambiguous. Here that simpler choice would couple “Replica dedupe hides missing data” to TSDB Ingest Node.	Costs schema migration, replication, backup, and shard ownership. Split storage only when access patterns, retention, or compliance boundaries materially differ. Measure the latency, saturation, lag, or mismatch at the WAL / Head Block boundary before extracting or scaling it.
Object Block Store	Owns authoritative versioned state or immutable evidence. In this design it participates in: Rule evaluators checkpoint durable timestamps and send deduplicated alert state.	Defines the source of truth and commit boundary. It prevents WAL / Head Block from also owning the contract and failure behavior of Compactor.	Let each service persist its own interpretation. That removes one clear commit boundary and makes retries, support, and reconciliation ambiguous. Here that simpler choice would couple “Rule evaluation misses interval during failover” to WAL / Head Block.	Costs schema migration, replication, backup, and shard ownership. Split storage only when access patterns, retention, or compliance boundaries materially differ. Measure the latency, saturation, lag, or mismatch at the Object Block Store boundary before extracting or scaling it.
Compactor	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Discover targets and scrape with bounded concurrency or accept authenticated remote write.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Object Block Store from also owning the contract and failure behavior of Query Frontend.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Unbounded label creates millions of series” to Object Block Store.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Compactor boundary before extracting or scaling it.
Query Frontend	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Normalize labels, reject cardinality violations, and append batches to a durable log/WAL.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Compactor from also owning the contract and failure behavior of Rule Evaluator.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Late sample targets compacted block” to Compactor.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Query Frontend boundary before extracting or scaling it.
Rule Evaluator	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Route each fingerprint to one ingest owner and compact head chunks into immutable blocks.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Query Frontend from also owning the contract and failure behavior of Alert Manager.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Query scans too many series” to Query Frontend.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Rule Evaluator boundary before extracting or scaling it.
Alert Manager	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Query frontend splits time/shards, deduplicates replicas, and caches partial results.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Rule Evaluator from also owning the contract and failure behavior of operators and repair tooling.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Replica dedupe hides missing data” to Rule Evaluator.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Alert Manager boundary before extracting or scaling it.

Decision 01

Chosen boundary vs simpler monolith

Prefer pull for service health and discovery; use agents or remote write for constrained environments.

A single deployable with one transactional store is the right starting point for a small metrics collection workload. The proposed boundary becomes justified when prefer pull for service health and discovery; use agents or remote write for constrained environments. Separating it contains the named failure “Unbounded label creates millions of series” without placing unrelated work on the critical path.

Decision 02

Correctness vs availability

Compress timestamp/value deltas, partition by series fingerprint, and downsample older data.

The design coordinates only where each accepted metrics collection operation has one stable result and retries do not repeat its business effect. Everywhere else it accepts bounded staleness or asynchronous propagation. The alternative—strong coordination for every projection—simplifies the mental model but increases tail latency and turns partitions into broad write unavailability.

Decision 03

Operational cost and evolution

Budget label cardinality explicitly; one unbounded user_id label can overwhelm the system.

This choice adds dashboards, ownership, schema compatibility, capacity planning, and recovery runbooks. Start with the minimum topology that supports “Route each fingerprint to one ingest owner and compact head chunks into immutable blocks.” and extract another service only when measured throughput, blast radius, data lifecycle, or team ownership changes independently.

Why this design: the synchronous path contains only the decisions required to protect each accepted metrics collection operation has one stable result and retries do not repeat its business effect. Durable projections, notifications, indexing, analytics, and repair are isolated so their load or failure cannot corrupt the acknowledged operation.

Correctness boundary

Prefer pull for service health and discovery; use agents or remote write for constrained environments.

Scale boundary

Compress timestamp/value deltas, partition by series fingerprint, and downsample older data.

Operational boundary

Budget label cardinality explicitly; one unbounded user_id label can overwhelm the system.

↑ top

5. Critical flow and algorithm

Discover targets and scrape with bounded concurrency or accept authenticated remote write.

Normalize labels, reject cardinality violations, and append batches to a durable log/WAL.

Route each fingerprint to one ingest owner and compact head chunks into immutable blocks.

Query frontend splits time/shards, deduplicates replicas, and caches partial results.

Rule evaluators checkpoint durable timestamps and send deduplicated alert state.

Retry-safe mutation skeleton

identity = authenticate(request)
prior = idempotency.get(identity, request.key)
if prior: return prior.result

current = authoritative_store.read(partition_key)
next_state, events = domain.transition(current, request)
authoritative_store.compare_and_set(current.version, next_state)
idempotency.commit(request.key, next_state.result)
outbox.append(events)
return next_state.result

Exactly-once business effect: the network can duplicate requests. Stable identities, conditional state transitions, immutable event IDs, and reconciliation make the user-visible effect happen once.

↑ top

6. Caching, hot keys, and backpressure

Cache policy

Add a cache only for a measured repeated read; do not place mutable correctness state in a disposable layer.

Hot partition or tenant

Measure per-key and per-tenant skew. Replicate safe reads, isolate the offender, salt only when ordering permits, and enforce budgets.

Backpressure

Bound queues, buffers, connection pools, and concurrency. Shed low-priority work and scale from lag/oldest-item age rather than CPU alone.

Stampede and retry storms

Use single-flight/coalescing, jittered expiry, bounded retries, exponential backoff, and circuit breakers so one dependency failure does not multiply traffic.

Cache outage rule: never allow every miss to fall through simultaneously. Use local protection, request coalescing, circuit breakers, bounded pools, and controlled fallback.

↑ top

7. Partitioning and consistency

partition = stable_hash(primary_or_aggregate_key) → shard owner
write = conditional_update(expected_version)
read_projection = versioned and rebuildable

Strong boundary

each accepted metrics collection operation has one stable result and retries do not repeat its business effect. Use a uniqueness constraint, serializable/conditional write, expected version, or fencing token at this boundary.

Eventual boundary

Analytics, notifications, search, read models, and operational dashboards may lag when freshness is observable and replay exists.

Operation	Consistency	Reason
Authoritative mutation	Conditional / serializable	Protect uniqueness, ownership, money, inventory, order, or legal state transition.
Read after own write	Primary, session token, or cache warm	The caller must observe the result it just committed.
Ordinary read projection	Bounded stale where safe	Replicas and caches improve latency when version/freshness is visible.
Analytics and operations	Eventual + replay	Freshness is less important than isolating the online path.

Route every mutation for one aggregate to one owner/partition epoch.
Use cursors rather than offsets for changing ordered collections.
Retain tombstones until replication and repair windows close.
Move shard ownership through versioned maps and dual-read verification.

↑ top

8. Failure drills and recovery

Failure	Runtime behavior	Recovery proof
Unbounded label creates millions of series	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Late sample targets compacted block	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Query scans too many series	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Replica dedupe hides missing data	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Rule evaluation misses interval during failover	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.

Recovery checkpoint

# Recovery checkpoint
scenario: Unbounded label creates millions of series
operation_id: op_01J...
aggregate_key: metrics-system_123
last_durable_version: 18
projection_checkpoint: partition-07@offset-482901
retry_policy: same_identity / bounded_backoff
reconciliation_status: pending
owner: Alert Manager

Every dependency needs a timeout, retry budget, circuit policy, and explicit fail-open, fail-closed, stale, or read-only mode. “Highly available” is incomplete until degraded behavior is named.

↑ top

9. Security, regions, and observability

Security and abuse

Authenticate at ingress and authorize the exact resource/tenant on every operation.
Encrypt transport and durable data; tokenize or isolate sensitive material.
Rate-limit by tenant, actor, resource, and network; detect enumeration and anomalous fan-out.
Redact secrets and personal data before logs or analytical storage.
Audit privileged mutations, support access, policy changes, and forced recovery.

Signals that prove correctness

User-visible success rate and p50/p95/p99 latency.
Timeouts, circuit state, saturation, and replica health by dependency.
Queue lag, oldest-item age, retries, and dead-letter growth.
Conflict, duplicate-suppression, and reconciliation mismatch counts.
Business drill: Unbounded label creates millions of series — alert before the user-visible invariant is violated.

Regional evolution

Begin multi-zone with tested failover, backups, and reconciliation.
Add global stateless ingress and regional read projections.
Keep one write owner per aggregate or partition where coordination is expensive.
Move ownership with epochs/fencing; avoid silent simultaneous writers.
Use expand/contract schemas, canaries, shadow reads, and reversible cutovers.

↑ top

10. Interviewer follow-ups

Push or pull—which and why?

Prefer pull for service health and discovery; use agents or remote write for constrained environments. Walk the concrete lifecycle step “Discover targets and scrape with bounded concurrency or accept authenticated remote write.” Then test it against “Unbounded label creates millions of series”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How do recording rules help?

Compress timestamp/value deltas, partition by series fingerprint, and downsample older data. Walk the concrete lifecycle step “Normalize labels, reject cardinality violations, and append batches to a durable log/WAL.” Then test it against “Late sample targets compacted block”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How do you handle late samples?

Budget label cardinality explicitly; one unbounded user_id label can overwhelm the system. Walk the concrete lifecycle step “Route each fingerprint to one ingest owner and compact head chunks into immutable blocks.” Then test it against “Query scans too many series”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How are alerts made highly available?

Prefer pull for service health and discovery; use agents or remote write for constrained environments. Walk the concrete lifecycle step “Query frontend splits time/shards, deduplicates replicas, and caches partial results.” Then test it against “Replica dedupe hides missing data”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How do you federate regions?

Compress timestamp/value deltas, partition by series fingerprint, and downsample older data. Walk the concrete lifecycle step “Rule evaluators checkpoint durable timestamps and send deduplicated alert state.” Then test it against “Rule evaluation misses interval during failover”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

Answer pattern: state the invariant, choose a mechanism, walk one failure, name the recovery signal, and explain the cost of the alternative.

↑ top

11. How to present this in 45 minutes

Time	What to cover
0–5 min	Clarify scope, workload, SLOs, and the correctness invariant.
5–10 min	Estimate throughput/storage and define concrete APIs and records.
10–20 min	Draw online, durable-write, and asynchronous/repair paths.
20–30 min	Explain the exact lifecycle, partition key, and concurrency control.
30–38 min	Discuss caching, hot keys, replication, and backpressure.
38–43 min	Run failure drills, security, observability, and regional evolution.
43–45 min	Summarize the five most important choices and trade-offs.

Final decision summary

Invariant	each accepted metrics collection operation has one stable result and retries do not repeat its business effect.
Partitioning	Stable aggregate/key hash with versioned ownership and measured hot-key isolation.
Commit	Expected version or uniqueness at the authoritative state boundary.
Async work	Durable outbox/log, idempotent consumers, checkpoints, and replay.
Recovery	Same operation identity, durable truth lookup, and business reconciliation.

Closing summary

CLOSING “The design protects each accepted metrics collection operation has one stable result and retries do not repeat its business effect. The online path stays bounded, authoritative state is versioned and partitioned, and asynchronous work is replayable and observable.”

↑ top

Instrumented Service

Scraper / Agent

Service Discovery

Remote Write Gateway

Complete System Design Tutorial

1. Requirements and scope

Functional requirements

Non-functional requirements

2. Capacity assumptions

3. API and data model

Authoritative records and indexes

Command and event envelopes

4. Why this architecture and what it costs

Chosen boundary vs simpler monolith

Correctness vs availability

Operational cost and evolution

5. Critical flow and algorithm

Retry-safe mutation skeleton

6. Caching, hot keys, and backpressure

7. Partitioning and consistency

Strong boundary

Eventual boundary

8. Failure drills and recovery

Recovery checkpoint

9. Security, regions, and observability

Security and abuse

Signals that prove correctness

Regional evolution

10. Interviewer follow-ups

11. How to present this in 45 minutes

Final decision summary

Closing summary